Information Retrieval Meets Information

Visualization PROMISE Winter School

2012 Zinal Switzerland January 23 27
2012 Revised Tutorial Lectures 1st
Edition Alan Dix (Auth.)
Visit to download the full and correct content document:
https://textbookfull.com/product/information-retrieval-meets-information-visualization-p
romise-winter-school-2012-zinal-switzerland-january-23-27-2012-revised-tutorial-lectu
res-1st-edition-alan-dix-auth/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Information Retrieval: 24th China Conference, CCIR

2018, Guilin, China, September 27–29, 2018, Proceedings
Shichao Zhang

https://textbookfull.com/product/information-retrieval-24th-
china-conference-ccir-2018-guilin-china-
september-27-29-2018-proceedings-shichao-zhang/

Introduction to Information Retrieval Manning

https://textbookfull.com/product/introduction-to-information-
retrieval-manning/

Complex Sciences Second International Conference

COMPLEX 2012 Santa Fe NM USA December 5 7 2012 Revised
Selected Papers 1st Edition Matthew Antognoli

https://textbookfull.com/product/complex-sciences-second-
international-conference-complex-2012-santa-fe-nm-usa-
december-5-7-2012-revised-selected-papers-1st-edition-matthew-
antognoli/

Mobile Information Retrieval 1st Edition Prof. Fabio

Crestani

https://textbookfull.com/product/mobile-information-
retrieval-1st-edition-prof-fabio-crestani/
Public Key Infrastructures Services and Applications
9th European Workshop EuroPKI 2012 Pisa Italy September
13 14 2012 Revised Selected Papers 1st Edition Masaya
Yasuda
https://textbookfull.com/product/public-key-infrastructures-
services-and-applications-9th-european-workshop-
europki-2012-pisa-italy-september-13-14-2012-revised-selected-
papers-1st-edition-masaya-yasuda/

Information Security and Cryptology ICISC 2013 16th

International Conference Seoul Korea November 27 29
2013 Revised Selected Papers 1st Edition Hyang-Sook Lee

https://textbookfull.com/product/information-security-and-
cryptology-icisc-2013-16th-international-conference-seoul-korea-
november-27-29-2013-revised-selected-papers-1st-edition-hyang-
sook-lee/

Engineering Trustworthy Software Systems: 5th

International School, SETSS 2019, Chongqing, China,
April 21–27, 2019, Tutorial Lectures Jonathan P. Bowen

https://textbookfull.com/product/engineering-trustworthy-
software-systems-5th-international-school-setss-2019-chongqing-
china-april-21-27-2019-tutorial-lectures-jonathan-p-bowen/

Web Information Systems Engineering: WISE 2019

Workshop, Demo, and Tutorial, Hong Kong and Macau,
China, January 19–22, 2020, Revised Selected Papers
Leong Hou U
https://textbookfull.com/product/web-information-systems-
engineering-wise-2019-workshop-demo-and-tutorial-hong-kong-and-
macau-china-january-19-22-2020-revised-selected-papers-leong-
hou-u/

Emerging Ideas on Information Filtering and Retrieval:

DART 2013: Revised and Invited Papers 1st Edition
Cristian Lai

https://textbookfull.com/product/emerging-ideas-on-information-
filtering-and-retrieval-dart-2013-revised-and-invited-papers-1st-
edition-cristian-lai/
Lecture Notes in Computer Science 7757
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen

Editorial Board
David Hutchison
Lancaster University, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Alfred Kobsa
University of California, Irvine, CA, USA
Friedemann Mattern
ETH Zurich, Switzerland
John C. Mitchell
Stanford University, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
Oscar Nierstrasz
University of Bern, Switzerland
C. Pandu Rangan
Indian Institute of Technology, Madras, India
Bernhard Steffen
TU Dortmund University, Germany
Madhu Sudan
Microsoft Research, Cambridge, MA, USA
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbruecken, Germany
Maristella Agosti Nicola Ferro Pamela Forner
Henning Müller Giuseppe Santucci (Eds.)

Information Retrieval Meets

Information Visualization
PROMISE Winter School 2012
Zinal, Switzerland, January 23-27, 2012
Revised Tutorial Lectures

13
Volume Editors

Maristella Agosti
Nicola Ferro
University of Padua
Department of Information Engineering
Via Gradenigo 6/a, 35131 Padua, Italy
E-mail: maristella.agosti@unipd.it, ferro@dei.unipd.it

Pamela Forner
Center for the Evaluation of Language
and Communication Technologies (CELCT)
Via alla Cascata 56/c, 38123 Povo, TN, Italy
E-mail: forner@celct.it

Henning Müller
University of Applied Sciences Western Switzerland
TechnoArk 3, 3960 Sierre, Switzerland
E-mail: henning.mueller@hevs.ch

Giuseppe Santucci
Sapienza University of Rome
Department of Computer, Control
and Management Engineering Antonio Ruberti
Via Ariosto 25, 00185 Rome, Italy
E-mail: santucci@dis.uniroma1.it

ISSN 0302-9743 e-ISSN 1611-3349

ISBN 978-3-642-36414-3 e-ISBN 978-3-642-36415-0
DOI 10.1007/978-3-642-36415-0
Springer Heidelberg Dordrecht London New York

Library of Congress Control Number: 2013930306

CR Subject Classification (1998): H.3.1, H.3.3-5, H.3.7, H.4.1, H.5.1-3

LNCS Sublibrary: SL 3 – Information Systems and Application, incl. Internet/Web

and HCI
© Springer-Verlag Berlin Heidelberg 2013

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting,
reproduction on microfilms or in any other way, and storage in data banks. Duplication of this publication
or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965,
in ist current version, and permission for use must always be obtained from Springer. Violations are liable
to prosecution under the German Copyright Law.
The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply,
even in the absence of a specific statement, that such names are exempt from the relevant protective laws
and regulations and therefore free for general use.
Typesetting: Camera-ready by author, data conversion by Scientific Publishing Services, Chennai, India
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)
 Preface

In the context of the European Union (EU)-funded research project PROMISE

(Participative Research labOratory for Multimedia and Multilingual Information
Systems Evaluation), a winter school was organized in the small ski resort of
Zinal, Valais, Switzerland during January 23-27, 20121.
PROMISE aims at advancing the experimental evaluation of complex mul-
timedia and multilingual information systems in order to support individuals,
commercial entities, and communities who design, develop, employ, and improve
such complex systems. The overall goal of PROMISE is to deliver a unified envi-
ronment collecting data, knowledge, tools, methodologies, and to help the user
community which is involved in the experimental evaluation.
The title of the winter school was From Information Retrieval to Information
Visualization and the goal was to bring together these two research domains
that are currently quite separated but have an important potential to help each
other in advancing the fields. Indeed, the school was attended by participants
who came from one domain or the other and offered them the possibility of
starting to acquire cross-disciplinary competencies. Interestingly enough, the
school turned out to be a brainstorming and discussion opportunity also for the
lecturers, since they had the chance to meet colleagues from a quite different field
with their own perspectives on a ground of shared topics and issues, such as how
to envision models and design systems around user needs, how to consider the
user interaction and context, how to conduct evaluation, and so on.
In all, 17 high-quality lecturers from academia and industry were invited to
speak on a large variety of topics from introduction talks to hot topics such
as crowd sourcing and social media; 62 participants from 25 countries and four
organizers followed the courses and helped to create lively discussions and an
open atmosphere with many questions. Most of the speakers stayed for the entire
week and enriched the discussions as well.
All participants had the possibility to present their own work with a poster
during the first day of the winter school at the evening welcome reception that
started many discussions among the participants. An evaluation and selection
of the posters was performed and a symbolic best poster award – a bottle of
excellent Swiss wine – was given to the three best posters.
The fact that the participants remained close together during the five days of
the winter school and had many possibilities to meet with the other participants
and the lecturers gave rise to many discussions and to a stimulating environment
for both the participants and the lecturers.

1
http://www.promise-noe.eu/events/winter-school-2012/
VI Preface

Altogether the PROMISE winter school can be seen as a great success in

connecting two research domains and allowing a large number of participants
to get in contact with high-quality lecturers and give them hopefully a better
view of the research domains and also on the ways in which they can evaluate
their own research and profit from tools of visualization that are available. Most
participants gave very positive feedback, and hopefully the proceedings of the
winter school will also help to record the main outcomes of the winter school for
the future and for those persons who could unfortunately not participate.
An analysis of the evaluation forms after the winter school highlighted that
most students very much enjoyed (more than 90% of the participants) the win-
ter school and the atmosphere among the participants and the lecturers. Most
presentations were enjoyed (about 90% of the participants); but sometimes the
introductory presentations were regarded as too simple for a majority of PhD
students. The students were generally interested in the different topics offered by
the school (about 90% of the participants). The setting in a remote alpine valley
was very much appreciated but the lack of a professional conference room and of
limited infrastructures such as for lunch time were also regarded as problematic.
Lectures of 90 minutes were regarded as too long, and perhaps short breaks after
45 minutes would have been a better option.

Acknowledgement
We would like to thank all the lecturers and participants to the PROMISE
Winter School who gave an extremely valuable contribution and made the school
a success.
We would like to thank those institutions and individuals who have made
this school possible: University of Applied Sciences Western Switzerland and
Conférence Universitaire de Suisse Occidentale (CUSO).
The PROMISE Winter School has been supported by: the PROMISE2 net-
work of excellence (contract n. 258191), the Khresmoi3 (contract n. 257528)
and CULTURA4 (contract n. 269973) projects, as part of the 7th Framework
Program of the European Commission, and by the ELIAS5 research networking
programme of the European Science Foundation.

December 2012 Maristella Agosti

Nicola Ferro
Pamela Forner
Henning Müller
Giuseppe Santucci

2
http://www.promise-noe.eu/
3
http://www.khresmoi.eu/
4
http://www.cultura-strep.eu/
5
http://www.elias-network.eu/
 Organization

The 2012 PROMISE Winter School was organized by the University of Applied
Sciences Western Switzerland, Sierre (HES-SO), together with Sapienza, Uni-
versity of Rome, Italy, and the University of Padua, Italy.

General Chair
Tiziana Catarci Sapienza, University of Rome, Italy

Program Committee
Maristella Agosti University of Padua, Italy
Nicola Ferro University of Padua, Italy
Henning Müller University of Applied Sciences Western
Switzerland, Switzerland
Giuseppe Santucci Sapienza, University of Rome, Italy

Local Organizing Committee

Alexandre Cotting University of Applied Sciences Western
Switzerland, Switzerland
Henning Müller University of Applied Sciences Western
Switzerland, Switzerland

Publicity Committee
Pamela Forner Centre for the Evaluation of Language and
Communication Technologies (CELCT),
Italy
Hélène Mazo Evaluations and Language resources
Distribution Agency (ELDA), France
 Table of Contents

Introduction to Information Visualisation . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Alan Dix

Principles for Human-Centred Design of IR Interfaces . . . . . . . . . . . . . . . . 28

Maria Francesca Costabile and Paolo Buono

Human-Computer Interaction View on Information Retrieval

Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Tiziana Catarci and Stephen Kimani

User-Oriented Information Retrieval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

Elaine G. Toms

User-Oriented Evaluation in IR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Kalervo Järvelin

Multimedia Information Retrieval in a Social Context . . . . . . . . . . . . . . . . 92

Stéphane Marchand-Maillet

TREC-Style Evaluations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Donna Harman

Visual Analytics and Information Retrieval . . . . . . . . . . . . . . . . . . . . . . . . . . 116

Giuseppe Santucci

An Introduction to Crowdsourcing for Language and Multimedia

Technology Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Gareth J.F. Jones

Medical (Visual) Information Retrieval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

Henning Müller

Is Visualization Usable for Displaying Web Search Results

in an Exploratory Search Context? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
Aline Crédeville and Dominic Forest

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

 Introduction to Information Visualisation

Alan Dix1,2
1
Talis, 43 Temple Row, Birmingham, B2 5LS, UK
2
University of Birmingham, School of Computer Science, Birmingham, UK
alan@hcibook.com
http://alandix.com/academic/teaching/Promise2012/

Abstract. This is a short introduction to information visualisation, which is

increasingly important in many fields as information expands faster than our
ability to comprehend it. Visualisation makes data easier to understand through
direct sensory experience (usually visual), as opposed to more linguistic/logical
reasoning. This chapter examines reasons for using information visualisation
both for professional data analysts and also end-users. It will also look at some
of the history of visualisation (going back 4,500 years), classic examples of
information visualisations, and some current challenges for visualisation
research and practice. Design of effective visualisation requires an appreciation
of human perceptual, cognitive and also organisational and social factors, and
the chapter discusses some of these factors and the design issues and principles
arising from them.

Keywords: information visualisation, human–computer interaction, HCI, visual

analytics.

1 Introduction

Information assails us in business, in science, in government and in day-to-day life,

from the processing of massive scientific streams at CERN to sentiment analysis of
millions of Twitter messages to gauge the popularity of a party, or to locate power
outages. Information retrieval is about selecting out of this morass of data, relevant
documents, images, and audio. Visualisation is about helping people make sense of
either the original data sources or the subsets of data obtained through information
retrieval. Both can operate independently, but also they have great power together.
This chapter is a short introduction to information visualisation. In it we will look
at a number of areas. First, in the next section, we will look at the definition and
scope of visualisation in general and information visualisation in particular. Most
critically, despite the term being 'visualisation', it may in fact involve other senses and
is centrally about the use of these senses to make sense of data. Visualisation has
various purposes and users; section 3 considers these, in particular the different ways
in which information visualisation is used by data analysts compared with data
consumers (whether a company CEO or newspaper reader).
While visualisation seems like a modern phenomenon, and indeed interactive
computer visualisation is comparatively recent, in fact the roots of static visualisation

M. Agosti et al. (Eds.): PROMISE Winter School 2012, LNCS 7757, pp. 1–27, 2013.
© Springer-Verlag Berlin Heidelberg 2013
2 A. Dix

can be traced back at least 4500 years. Section 4 presents a brief history of
visualisation from Mesopotamian financial tables and 10th century line graphs to
current spreadsheet graphics, data journalism and visual analytics. This is followed in
section 5, by an overview of some of the kinds of visualisation that might be
particularly useful in the context of information retrieval.
The chapter concludes with a discussion of some of the human-centred design
principles that can be applied to visualisation choice and creation. We will consider
detailed design issues; in particular the way interaction can soften the trade-offs that
are inherent in making (static) visualisation choices. However, we will also consider
the way visualisation (and for that matter information retrieval) fits into a larger social
and organisational context.

2 What Is (Information) Visualisation

Defining Visualisation
Visualisation is perhaps easier to recognise than define. In his textbook "Information
Visualisation" [1], Bob Spence refers to the dictionary definition:
visualize: to form a mental mode or mental image of something [1]
He emphasises that visualisation is critically about insight, what happens in your
head, not a computer. Often the most powerful mental images are formed from words
alone, but that would not correspond to the common notion of visualisation, which is
often about the design of media (computer, paper) to help people.
So, for this chapter we shall adopt a slightly different definition of (information)
visualisation:
making data easier to understand using direct sensory experience
Note this is still about insight and understanding, but also about the perception
('sensory experience') and deliberate design ('making').
Note also that this definition says 'sensory', not simply 'visual', as the inner
visualisation that makes you say "I see" can also be engendered by other senses.
Although less common, you can have aural and tactile 'visualisation' – think of the
click of a Geiger counter – faster clicks mean more radiation. These non-visual forms
are particularly valuable for those with visual disability, but also in contexts when the
eyes need to be elsewhere, for example while flying a plane. This all said, the vast
majority of visualisation is, as the name suggests, visual. The visual cortex accounts
for around 50% of our brain, and so it makes sense to use it.
Also note that the word 'direct' is in the definition to exclude purely rich textual
descriptions, no matter how sharply they focus the mind. Although you use your eyes
to read words or even tables of numbers, they are processed linguistically and
logically, rather than the more instant feeling you get when you see a rising graph.
With many caveats to beware of pseudo-science, you can think of this as a form of
left brain / right brain distinction. Not that one is better than the other. In statistics,
one is taught never to start by calculating means, t-tests, etc., but instead always to
 Introduction to Information Visualisation 3

start off by drawing graphs, trying to get a feel for the data (very right brain). Thiings
are often obvious by simply glancing at a graph. However, having got an ideaa of
what one thinks is true of the
t data, one does not trust that intuition, but then starts to
calculate the statistics (very
y left brain) to verify the insight. The two work togetherr.

Visualising Numbers
This does not mean that teext and numbers are not an integral part of visualisatiion.
Good layout can create direct visual (or other sensory) impressions. Look at the ttwo
columns of numbers in Fig gure 1. In each column try to see, as quickly as possibble,
which is the biggest numb ber. This is harder in the left-hand column than the rigght-
hand one. This is because the numbers in the right-hand column have their decim mal
points aligned, so the biggeest numbers are also the ones that stick out furthest to the
left of the decimal point. Efffectively the line of figures acts like a miniature bar graaph.

Fig. 1. Visualising in numbers

Of course one may no ot care about the biggest numbers, in which case the
alignment doesn't matter. As a first heuristic for information visualisation: "thhink
purpose" – work out what you y want the viewer to be able to do with the visualisattion
and use that to determine th he form.
A more complex form of o visualisation, where the numbers are still central, cann be
T is like a spreadsheet except that columns and rows can
found in Table Lens [2]. This
be collapsed down to a few w pixels each. Where a cell is not collapsed the numbers can
be read (and will be aligned d properly!). However, when the height of the cell becom mes
too small to show the numb ber it is reduced to a line of pixels so that the column eends
up a sort of mini-histogramm. By sorting the middle column, it is immediately obviious
that the column to the left isi correlated to some extent with it, but that the right hhand
column far less so. When the t column width is collapsed the histogram becomes eeven
more miniature (columns 2 and 3 in figure 2), giving less detail, but still allowingg an
at a glance view of the tiny columns.
This is also an example of a general visualisation technique called 'focus+conttext'
or fisheye view [3]. The expanded
e cells allow one to look at certain values in deetail
(the focus), whilst the collaapsed cells allow one to get an idea of how that fits into the
big picture (the context).
4 A. Dix

Fig. 2. Table Lens [2]

Information Visualisation
n
The term 'information' visualisation, as opposed to 'visualisation' in general, is usually
used to contrast it with 'scieentific' visualisation. In science there are many phenom
mena
that have a direct connectio on to the physical world, but are in some way invisible, for
example the airflow around d an aircraft wing. This scientific data is often in the foorm
of fields of numbers of vecttors defined continuously over a 2D or 3D space.
In contrast, information visualisation is often concerned with data sometimes m more
complex structurally, but almost always discrete: hierarchies, tables, point ddata.
Furthermore, the data of infformation visualisation often includes categorical data ((e.g.
gender male/female) as welll as continuous data (e.g. height).
The two are not entirely y distinct, for example, geographical information systeems
(GIS) involve data over 2D maps. Methods used to display regional petrol consumpttion
for marketing purposes will not be so different from those showing average temperatture
patterns for climate modellin ng, and of course one might want to use both these data sets
to understand patterns of gloobal warming.

3 Why Use Visuallisation and Who Is It for?

When creating visualisation ns there are two kinds of target audience.

First there is the data analyst,
a whose job it is to sift through data whether the
academic interpreting exp perimental results, the forensic accountant looking for
anomalies in a bank's accou unts, the city planner working out the best route for a nnew
cycleway, or the intelligencce officer piecing together emails, tweets and passport ddata
to prevent a terrorist attack..
The other group is the eventual data consumer, the client, audience, newspaaper
reader, or the CEO. These may range from a time-strapped manager to an illiterrate
peasant, but have in comm mon that they are not experts at data analysis, and may not
even be highly numerate beeyond what they recall of basic school mathematics.
 Introduction to Information Visualisation 5

Given you are reading this

t book, it is likely that you have more in common w with
the first group, the scientist, statistician or professional, than the second. This meeans
you need to work harder to o design visualisations for the data-consumer, as you w will
not have as intuitive a grasp
p of what is good.
For each of these two groups we'll look at reasons why you may want to use
visualisations.

For the Data Consumer

For the data consumer, thee focus usually needs to be on simple, well understoood,
representations, that can bee grasped at first time of looking. Sometimes, for exampple,
when using a visualisation n as part of a presentation, it is possible to introducee an
audience to a more compleex graphic, but often the visualisation has to work at ffirst
glance or not at all.
There are two main reaasons for providing visualisations to the data consum mer:
understanding and rhetoric.
understanding – This is when we want to help others see what the analyst has
already seen. For examplee, as part of teaching a course on mobile internet, you m may
want a graph to show the nu umber of people accessing the internet via a mobile plottted
against time. For the gen neral public, graphs are often augmented with text and
graphics to form 'infographics'. This is partly to make the visualisations more visually
appealing (if the readers doo not look at the visualisation they will learn nothing), and
partly to point out particullar features. For example, the page in figure 3, from the
Guardian Datablog [4], sh hows the UK budget deficit from 1979 (the start of the
Thatcher administration) un ntil 2011. The colours denote the dominant party in pow wer
(blue=Conservative, red=Labour), the pictures at the top are the various Chancelllors
of the Exchequer at the timee and actual numbers included in the figure.

F 3. UK Deficit and Borrowing [4]

Fig.
6 A. Dix

rhetoric – Visualisation can also be used to persuade readers of a particular pooint,

whether valid or not. For example, every business plan includes a 'hockey sttick'
graph (see Fig 4.), that plo
ots projected users / income over time, starting slow, but
eventually taking off, showwing the potential investor that this is a good businesss in
which to invest. When we see graphs it is easy to be impressed, whether or not tthey
are comprehensible. They seem professional, scientific, and, sadly, often the lless
comprehensible they are, the
t more people are impressed (if they are difficult tthey
must be clever!). Rhetoricaal use of numbers or graphs can be misleading1, or cann be
used to convince others of the truth, whether in an academic paper, newspaper, paarty
political pamphlet, or PhD thesis.
t

Fig. 4. A 'hocckey stick' graph, as found in most business plans

It would be nice to say that the answer to rhetoric is facts, but sadly the worldd is
not like that. The good guy ys have to tell as good a story as the bad guys! Howevver,
data certainly helps. Note that
t as well as infographics, the Guardian Datablog alloows
the download of the raw data
d behind the visualisations; a part of the practice callled
'data journalism'. This meaans that a more experienced reader can download the ddata
and manipulate or visualisee it in any way that they wish, crucial for informed debbate
and democracy in an age wh here information is power.
Often small changes in a diagram can make a big difference to the lesson peoople
take away. Consider figuree 5, this shows UK deficit again (up means deficit, so bbad,
in this graph). However, un nlike figure 3, this has been corrected for GDP (figures aalso
obtained from the Guardian n Datablog [4, 6]), so that deficit is shown relative to the
size of the economy and hence its affordability (like looking at your credit card bill
relative to your income.) Inn both figures it is clear that the deficit shot up in 2009 w
with
the world credit crisis. Figu
ure 3 also shows that with successive governments defiicits
have risen and fallen, with even the occasional foray into surplus. However, the grraph
also suggests a tendency fo or it to increase with time. Correcting for GDP, makees it
clear that the overall levell in real terms has remained pretty steady, with the last
administration actually sub bstantially lower than the long term average (2% G GDP
compared with 3%), counteer to the popular narrative of all political parties.

1
As Winston Churchill said, "the
" only statistics you can trust are those you falsified yourself",
or as has been attributed to Disraeli
D and others "there are three kinds of lies: lies, damned lies,
and statistics" [5].
 Introduction to Information Visualisation 7

Fig. 5. UK
U Deficit relative to GDP (data from [4, 6])

For the Data Analyst

While many of the same lesssons are true for the experienced data analyst, there is aalso
the opportunity for training g, or growing experience, so visualisations can afford too be
more complex and powerfful, potentially including novel techniques. Visualisattion
must still in a sense be 'first glance', as the purpose is to use the power of our senssory
perception; if you have to spend too long figuring out what a visualisation meaans,
then probably you are betteer looking straight at the numbers. However, this can bbe a
'first glance' after you have extensive experience and training. For example, engineeers
have many specialised graaphs used to understand fluid flows, electric circuitss or
cybernetic systems. Some take several years of training during undergraduate studyy to
master, but once mastered offer
o an instant overview.
c also consider two kinds of purpose: understanding and
For the data analyst we can
exploration.
understanding – Like the data consumer, the analyst may in a sense 'knnow
something' at least in abstrract terms, but wish to make it salient to themselves. For
example, we may be expecting a power law, so plot points on a log-log scale. For the
academic, the graphics used to help oneself understand may well be similar to thhose
published in an article as th he audiences (oneself and other scientists) are similar. For
example, figure 6 shows a box plot from the famous neutrinos faster than light paaper
[7]. For publication purposes the aim is to help others see (or convince otherss to
believe) what you have seen s in the paper. For the scientist who performed the
experiment, the purpose is to confirm/disconfirm hypotheses, highlight exceptionss or
outliers.
exploration – The otherr purpose for the data analyst is to find new things tthat
haven't even been considereed before. This may be a scientist encountering a new kkind
of data or new phenomeno on, or an intelligence officer seeking patterns amongst the
chaos of billions of intercep pted emails. In the previous cases, the design or selecttion
of visualisation is driven by b the desire to expose and clarify a previously knoown
pattern. Here the aim is to seek
s the unknown, indeed maybe deliberately try to dessign
visualisations that avoid th he obvious, exploring new angles (maybe literally iin a
multidimensional plot!). Typically
T this may involve flipping between different kiinds
of visualisation, each of wh hich may emphasise one aspect of the data, but hide othhers,
or maybe present several viisualisations at once (see Figure. 7).
8 A. Dix

Fig. 6. Box plot of Neutrino transit times [7]

Fig
g. 7. Multiple parallel visualisations [8]

The human sensory systeem is tuned to find patterns, and this is exploited to the full
nalysis, especially exploratory. However, we may also see
in visualisation for data an
patterns where there are no one. The more different ways you look at something, the
more likely one will appeear to have a pattern, purely by chance. Visualisatioons,
particularly those for exp ploratory analysis, therefore need to help the anaalyst
distinguish happenstance from real underlying patterns.

4 A (Very) Brief History

H of Visualisation

4.1 Static Visualisation

n (From 2500 BC to 1990 AD)
The computer-driven visu ualisations shown so far are comparatively recent, but
visualisations of various fo
orms date back many millennia. The Mesopotamian cclay
tablet in Figure 8 is arounnd 4500 years old, and contains a table of administrattive
information. We may think k bureaucracy is new, but the vast majority of early cclay
tablet writing is of an adm
ministrative / financial nature, often including simple tabbles
of numbers.
 Introduction to Information Visualisation 9

Fig. 8. Mesopotamian table on

o a clay tablet Fig. 9. 10th Century time line

Moving on 3500 years, Figure 9 shows an early line graph of solar, lunar and
planetary movements. Thee x-axis is days in the month and the lines track eeach
heavenly body, where the y-axis
y is their height in the sky. Figure 10 skips forwardd to
the 19th century and showss a visualisation of the Paris–Lyon train timetable, with the
x-axis hours of the day (fro om 6am to 6am the next day) and the y-axis showing the
distance along the route (P Paris at the top Lyon at the bottom). Fast trains stand out
clearly as the steeper lines.
These early visualisation ns were created painstakingly by hand, but with the advvent
of computing it became po ossible to create the same visualisations more quicklyy or
easily, or to create new on nes that would have been impossible before. In the eaarly
days this was done using veryv slow x-y pen plotters or character-graphics on a lline
printer, but now it is simply y a matter of selecting a few options in Excel and presssing
print!

Fig
g. 10. 1855 Paris-Lyon train timetable

This use of computers to create fixed visualisations whether on screen, on a w

web
page or PDF, or printed in
n a newspaper, is in many ways similar to the older haand-
drawn illustrations. These static visualisations are still of great importannce,
particularly when commu unicating with others. Furthermore, understanding the
10 A. Dix

effective design and qualities of static visualisations is an essential first step to

creating more complex interactive visualisations. For static visualisation, the core
texts are undoubtedly Tufte's beautifully illustrated books [9–11].

4.2 Interactive Visualisation

Examples of interactive visualisation can be traced back to early scanning vector

graphics displays, or the seaside information boards where tiny lights were
illuminated when you pressed buttons for different kinds of features. However, it was
in the early 1990s when growing graphics power made it possible, for the first time, to
create rich 3D graphics, complex visualisations and real-time interaction. This led to a
blossoming of information visualisation (and other graphics) research notably in the
groups at Xerox PARC and University of Maryland. Not all the ideas were good, just
like with gloriously multi-fonted documents during the desktop publication revolution
in the 1980s, there were many examples of gratuitous 3D, most of
which are deservedly forgotten. However, despite this, most of the core kinds of
visualisations in use today were introduced at that time (see selection in Figure 11),
several of which will be discussed in the next section.

Fig. 11. Interactive Visualisations from the early 1990s: clockwise from top left: Cone Trees
[12], TreeMaps [13], FilmFinder [14], Buttefly Browser [15], and Pixel Plotting [16] in centre
(note how use of 3D distorts text in Butterfly Browser)

4.3 Current Directions

We have already seen examples of data journalism where rich, but simple to
understand, infographics have made their way into mainstream media. Furthermore
the web has increased the public expectations of high quality, often interactive,
visualisations. These web visualisations are sometimes 'authored', that is created by
 Introduction to Information Visualisation 11

the individual or institution responsible for the article or blog. However, there are aalso
a number of data sharing anda analysis sites that make it easy to upload and visuaalise
your own data; figure 12 sh hows one example, IBM's "Many Eyes" [17]. Furtherm more
open data initiatives by go overnments and corporations across the world are makking
data on many aspects of liife available to all from the environment to employm ment,
crime to concert venues. Often this comes with an invitation to mashup and visuaalise
the data in citizens’ own intterfaces.

Fig. 12. IBM Many Eyes [17]

Various factors including g eScience and the web itself have led to an increase in the
volume of available data,, from the scientific data streams of ozone monitorring
stations, to the trivia of Twitter streams. Analysing this data has become big businness
and a major challenge. Happily,
H in parallel with the rise in big data there has
been a rise in processing power,
p both on the desktop and also in the cloud wherre it
is now relatively easy to spin up significant computational power as needed and
computational frameworkss, notably MapReduce [18], for dealing with distribuuted
computation on massive datta.
Where this computation meets visualisation, the nascent field of visual analyticcs is
growing [19, 20] (see the chapter later in this volume). This combines machhine
learning and other data pro ocessing algorithms with interactive visualisation to enaable
interactive problem solving g.

5 Classic Visualisa
ations for Information Retreival
There are at least as many
y kinds of visualisation as there are kinds of data. In this
section we will look at a few classes that are particularly relevant for informattion
retrieval.

5.1 Hierarchical Data

Trees, taxonomies and hieerarchies are perhaps the most ubiquitous form of ddata
structure after the humble table; we all encounter trees whether organisation chaarts,
biological taxonomies, (parts of) formal ontologies, XML data, or file systtem
hierarchies. After file browwser/outliner style textual layouts, the most common foorm
of tree visualisation is somee sort of box and line 'organisation' chart as in Figure 133.
12 A. Dix

Fig. 13. Simple tree visualisation

However, these simple visualisations

v tend to breakdown as trees get bigger. Figgure
14 shows a relatively smalll tree (part of a task structure), but even on this small trtree,
we can see a number of pro oblems of scale. The boxes are labelled, but it is often hhard
to fit the labels in the boxees without them overlapping. Furthermore, as one lookks at
lower levels of the tree, thee width grows very rapidly leading to extensive horizonntal
scrolling, and consequent loss of context. One option is to use 3D, and figuree 15
shows the Cone Tree [12 2], which lays out the tree nodes in rings verticallyy or
horizontally (Cam Tree) connected
c to their parents, creating 'cones'. When a uuser
selects a node the relevant ring
r swings round so that the selected node is in the cenntre
of attention.

Fig. 14. Simple tree visualisation

While the use of 3D ap pparently increases the amount of space available we still
view these in 2D, so the occclusion is still there, just more acceptable as we view iit as
the natural essence of the 3D world. Note how the shadows are used to help m make
sense of the trees in figurre 15. The vertical layout on the left of Figure 15 had
particular problems with teext labels, solved partially by the horizontal layout on the
right – small differences maatter in visualisation.
 Introduction to Information Visualisation 13

Fig. 15. Using 3D

D: the Cone Tree, vertical and horizontal variants [12]

Another approach is TreeMaps, which dissect and fill 2D space [13]. TreeMaps are
particularly suitable for treees where there is some notion of size or volume (e.g. ddisk
utilisation in a file system, staffing or spending in organisational units). The TreeM Map
divides the space horizontallly and vertically on successive steps, using the 'size' off the
subtree to determine the space
s allocated. In the end each smallest level has aarea
proportional to its size (see Fig. 16).

Fig. 16. TreeMap of two level

l hierarchical data: data on the left, TreeMap on the right

Fig. 17. TreeMap variants for images at leaves and large numbers of nodes [13]

Early versions of TreeeMaps applied the simple alternating horizontal/verttical

layout algorithm. Howeverr, later variants have divided space differently to avvoid
artefacts such as many thin
n rectangles when a node has a large number of childrren,
14 A. Dix

particularly important when displaying images in the leaf nodes (Figure 17, left). The
TreeMap is in many ways quite simple, but is one of the more heavily used 'complex'
visualisations, proving itself able to manage vast trees (Figure 17, right) and yet still
be relatively comprehensible.
Finally for here, although not the end of the visualisation of hierarchical data by
any means, are methods that distort space in order to show a tree. The most well
known (but not most well used or understood) of these is the Hyperbolic Browser
[21]. This began with the failure of simple circular layouts to deal with larger trees. If
a tree has a constant branching factor, say on average 3 nodes per parent, then the
number of subnodes increase exponentially at each level down: 3, 9, 27, 81, 243, 729,
... However, when we layout in a circle, then the circumference of successively larger
circles only grows linearly – there is never enough space!
Mathematicians deal with a kind of curved space called hyperbolic geometry. This
had theoretical beginnings, but now turns out to have applications in cosmological
physics. The important feature of hyperbolic space is that the circumference of
'circles' in this (rather odd) geometry increases exponentially with the diameter of the
circle – perfect for trees. Unfortunately we do not see in hyperbolic geometry, so this
is then projected back down into 2D (see Fig. 18) leading to an effect rather like
earlier Fish Eye visualisations [3].

Fig. 18. The Hyperbolic Browser visualising the web [21]

5.2 Clustered Data

Quite frequently in information retrieval there is no fixed structure, instead, we have
large sets of search results, with common attributes, but no given hierarchical
structure. Although there is no given structure, sometimes a form of structure is
induced using clustering, whether at a single level to give groups of related nodes, or
at multiple levels with clusters of clusters leading to tree structure.
Hierarchies are clearly centred on our linguistic/logical understanding of the world,
and to some extent need to be made more immediate to our sensory perception. In
contrast, clusters correspond closely to human perception, we see a group of sheep
and, without consciously thinking, "they are all similar" they become a flock in our
minds. However, the fuzzier concept has its own challenges.
 Introduction to Information Visualisation 15

Where the data is numerric and can be shown on some sort of scatter plot, there are
obvious ways to show a clu uster. Figure 19.a shows a group of elements that have bbeen
identified as a cluster. Th his might have been done by an automatic algorithm,, or
maybe by the user choosin ng or encircling elements interactively. If they are beeing
visualised on a 2D plane liike this, then we may visualise the cluster by showingg its
extent, perhaps by drawing g a border around the cluster (Fig. 19.a) or shading the
area included in the cluster (Fig. 19.b, shading). This is particularly appropriate w
when
there is an obvious region,, for example, if the user has lassoed the elements, or the
clustering algorithm createes a segmentation of space. Of course this may not bee an
ellipse as in Fig. 19, but perrhaps an area delimited by lines, or a more complex shaape,
and if there is not an obviouus region we might just use the convex hull of the pointss.

(a) scattered datta elements (b) show average or extent

F 19. Visualising numeric clusters

Fig.

The other obvious way to show a cluster is using some form of average positiion.
The large point in Figure 19.b
1 is not one of the original points, but an average off all
the points. Using an averag ge like this is an advantage if we want to reduce the cluutter
of the display [22], reduciing the number of points displayed by just showing the
centre point of each clusteer. Of course, this can be odd for some sorts of data, for
example, the (in)famous av verage of "2.2 children" per family in the 1970s. Howevver,
numbers at least admit this form of representation.
Things get far more diffiicult when the clusters represent non-numeric data, imagges,
text, or even categorical daata such as gender – 0.2 of a child is at least more eaasily
comprehensible than 55% female. Sometimes this kind of data is mapped into 2D
space, for example with multi-dimensional scaling, in which case this deriived
numeric data can be used to t display in the same way as numeric data. However, the
average value of such derrived statistics is likely to become increasingly hardd to
interpret, and you are still left with the problem of what details to show if the uuser
selects the 'average' elemennt.
Where the data is categ gorical there may be some attributes that are commonn to
most or all of the cluster, in
i which case these may be used, but for rich media: ttext,
images, sound, rather than n trying to compute some form of average, or generaated
archetype, it is often better to give one or more real examples.
These examples may bee deliberately chosen to be 'typical' using some meassure
(Fig. 20.a). For example, with
w text one may compute similarity measures basedd on
co-occurrence of words and a then use examples that are central based on tthis.
16 A. Dix

(a) show typicall elements (b) show spread of elements

Fig
g. 20. Visualising non-numeric clusters

Alternatively one may delib berately look for examples that are spread widely over the
elements in the cluster (Fiig. 20.b). When there is no systematic way of choossing
these, one can simply rando omly choose a number of examples.
This use of example dataa points can be seen in the Scatter-Gather browser, a classic
visualisation of clustering for
f text documents. Figure 21 shows the main window w of
the Scatter–Gather browserr (bottom left), which consists of two columns with ffive
regions in each. Each of th hese regions represents a cluster. The inset shows a cloose-
up of one of these cluster reepresentations.

Fig
g. 21. The Scatter–Gather Browser [23]

The Scatter–Gather brow wser is designed to help users find documents basedd on
hard to state or 'recognised when seen' criteria. The system starts with some collecttion
of documents, perhaps the entire
e library or perhaps the results of a keyword searchh. It
then uses an automatic algo orithm to cluster the documents into 10 clusters. The ussers
selects one or more clusterss that look interesting, the system gathers these into a nnew
set, and then repeats the process
p clustering the chosen document set into 10 nnew
clusters. Eventually, when the
t clusters are small enough, the user can swop to a m more
conventional view for final selection.
The problem here is thaat the original library may consist of many thousandss or
millions of documents so that the early clusters themselves consist of very laarge
sets. In the inset image in Figure 21 there are two main regions. On the top arre a
number of keywords. Thesee are commonly occurring words in the cluster – a form m of
 Introduction to Information Visualisation 17

the 'common attribute' visu

ualisation. Below that area are three individual documeents
represented by a short titlee or snippet; that is, a form of visualisation by choossing
typical elements.

5.3 Multi-attribute Datta

Data items have many attriibutes, for example a bibliographic record may have tiitle,
authors, date and place ofo publication, number of pages, references, citatioons,
keywords and taxonomic cllassification.
The earliest approach to
t such data was some form of boolean query usinng a
Fig. 22). This form of querying is still possible both for
command-line interface (F
traditional SQL databasess and more recent databases such as RDF data ussing
SPARQL or noSQL databases such as MongoDB's command line mode [24].

> new query

y
> type=‘journal’ and keyword=‘visualisation’
query processing complete - 2175 results
list all (Y/N)
> N

Fig. 22. Boolean queries at a command line

Of course ordinary userrs are not expected to use this form of query languaage,
however, the most common n interfaces are really only one step beyond, giving soome
sort of search form allowinng target values to be entered against different fields. For
example, Figure 23 is the Gmail
G advanced search form. Boolean queries dressed upp!

Fig. 23. Google Gmail search form (http://mail.google.com/)

More sophisticated inteerfaces allow faceted browsing. This is where seveeral

selection attributes are sho
own simultaneously then, as one makes selections agaainst
18 A. Dix

one, the options on otherss are narrowed correspondingly. For example, Figure 24
shows three document attriibutes: keywords, author and document type. The user has
selected 'interaction' and 'visualisation' from the keywords and 'journal' as the
document type. The count '173' shows the total number of documents satisfying bboth
– effectively documents whhere:

( "interaction" IN keywords OR "visualisation" IN keywords

s )
AND type = "journal"

Note that against each autthor and keyword is a count. This shows the numberr of
selected documents that alsso include the relevant attribute. Note that in the keywoords
list 39+157>173 as there arre 23 documents with both keywords.

Fig. 24. Faceted

F browsing, similar to HiBrowse [25]

There are many examplees of faceted browsing, both for conventional tabular ddata
and also alternative data in ncluding Semantic Web data [26]. The images above are
based on one of the earliestt, HiBrowse [25], which was used for various applicatiions
including large document reepositories and hotel selection.
The dynamic counts aree a critical feature of the HiBrowse. With early comm mand
line query interfaces it wass easy to refine a search and then find it ended up withh no
results. In Figure 24 it is ob
bvious that if you decide to choose the author 'smith', tthen
you will have no results – that
t is the counts give a sort of peek over the horizon as to
what will happen after you ur next interaction. This is a very powerful, but underussed,
visual interaction heuristic.

Fig. 25.
2 Dynamic filtering in FilmFinder [14]
 Introduction to Information Visualisation 19

In the above example, the interactions were discrete, selections of attributes leading
to updated values. However, sometimes interactions can be made more rapid and
continuous. Figure 25 shows two screenshots of FilmFinder [14], another early
faceted browsing interface. Within each screen there is a larger area to the left which
shows a scatter plot of films, coloured by genre and plotted against date (x-axis) and
popularity (y-axis). On the right are a number of sliders, which allow the setting of
maximum and minimum values for various attributes. As the user moves these sliders,
the points on the scatter graph are filtered in real time giving instant feedback. Note
that in the right-hand screen shot the filtering has reduced the number of points and so
the titles of the films are also shown.
Another example of faceted browsing is the Influence Explorer [27, 28]. This was
designed to allow exploration of complex engineering problems including simulations.
An example problem was light bulb design choices. There are various input
parameters that can be chosen (e.g. material, thickness and length of filament), and
various output measures (e.g. cost, lifetime). Large numbers of simulations are run to
create a large set of multi-dimensional data points, each corresponding to a single
simulation run. The engineer can then use dynamic sliders to either select 'input
parameter' ranges (e.g. choose range of thicknesses), or 'output' parameters (e.g.
maximum cost). So far, this is like the FilmFinder interface, except above each slider is
a small histogram showing the way the currently selected items (simulation runs) are
distributed over the relevant values. This is effectively like the counts in HiBrowse
making it possible to see whether the sliders are hovering near critical values.

Fig. 26. 'Peek over the horizon' histograms in Influence Explorer [27, 28]

5.4 Big Data

One of the trends noted in section 4.3 is the vast data sets that now need to be
analysed. Many visualisations fail when dealing with large data. Some problems are
computational, simply too many points to perform calculations on, especially for real
time interactive visualisation. Some problems are more intrinsic to the visualisation,
for example if there are too many points on a scatter plot it becomes unreadable, just
solid colour.
One approach is to simply use less space to visualise each item. Figure 27 shows
VisD, an example of pixel plotting [16], which uses a single pixel for each data
point and then packs these densely filling the available space. In Figure 27 the
20 A. Dix

pixels are plotted in circles starting in the centre and then spiralling outwards. Similar
techniques are also used for filling square areas. The colour represents a single
attribute of the data, and some other attribute is used to order the plotting. For
example, if the data is ordered by time then trends in the data would appear as
changes in the average colour between centre and periphery, and periodicity would
show up as segments or swirling patterns.

Fig. 27. Pixel plotting [16]

Pixel plotting allows 100s of thousands or millions of data points to be plotted on

an ordinary display, but still this does not help with many current datasets, for
example the tens of billions of web pages on a typical crawl.
For these vast datasets there needs to be some form of data reduction. This may
take the form of some sort of pre-programmed or emergent aggregation. For example,
detecting clusters and displaying the cluster averages, as described earlier – that is
visualising groups not individual elements. Example data points can also be used,
effectively reducing the number of points displayed. These examples might be
selected using some systematic technique, or simply using random sampling [29].

6 Designing for Visualisation

6.1 Perception and Purpose

As we have seen there are many different forms of visualisation. When choosing a
visualisation or designing a new one there are several factors to take into account:
visual ‘affordances’ – what we can see – Our eyes are better at some things than
others. For example, they are much better at discriminating levels of darkness, than
hues of colour, and are much better comparing lengths of lines when the lines share a
common base.
objectives, goals and tasks – what we need to see – Recall the lists of numbers in
Figure 1, if the purpose is to compare sizes or find the biggest/smallest, then aligning
the decimal points helps you to do this. If you can understand the purpose of a
visualisation, you are in a better position to ensure that the visual affordances make
this purpose achievable.
 Introduction to Information Visualisation 21

aesthetics – what we likel to see – Sometimes visualisations simply need too be

functional, but often they also need to be attractive. This is certainly true of the
infographics intended for public consumption. However, it is also true of professioonal
systems as we all work bettter when things are pleasant to look at.
These different visualisaation factors often conflict. For example, business repoorts
often use 3D charts as they y look impressive, even though it is usually harder to m
make
visual discrimination of (staatic) 3D images. Particularly common, and problematic are
3D pie charts. Pie charts arre difficult anyway as our eyes are not good at comparring
angles, but when the pie ch hart is put in perspective it becomes nearly impossiblee to
compare the resulting differrently shaped segments.
For some purposes the fu unctional aspects, the fit between perception and purposee, is
most important. For others, including the persuasive use of visualisation, the aestheetics
may be as or even more important. Furthermore there may be several intennded
purposes, which may each be optimally suited for different visualisations. As inn all
design the art is in choosing
g an appropriate trade-off between these conflicting goalss.

6.2 Interaction

One of the advantages of in nteraction is that it can relax some of the trade-offs intrinnsic
in visualisation by allowin ng some choices to be altered dynamically. This cann be
used in complex visualisattions, for example, the FilmFinder allowed the settingg of
parameter ranges for filterin ng which would have to be chosen beforehand and fixedd in
a static visualisation. Howeever, very simple visualisations can be made surprisinngly
powerful by just a little inteeraction.
As an example, let's con nsider the simple stacked histogram of fruit sales in Figgure
28. This representation is quite
q good at giving one a sense of the overall trend of the
total over time (the overalll height of the bars) and of the breakdown of fruits witthin
overall sales. It is also easyy to see the trend over time of apple sales, as they are the
bottom category. However it is very hard to see the trends of other fruits, for exampple,
are the sales of bananas inccreasing or decreasing? With a static stacked histogram, not
all fruits can be equally eaasy to visually analyse and so the designer needs to m make
choices and trade-offs.

Fig. 28. Stacked Histogram

22 A. Dix

However, if the histogrram is augmented with interaction, this trade-off cann be

relaxed. Figure 29 shows (aa static screen shot of) interactive stacked histograms (aalso
called 'dancing histogramss') [30]. Two very small interactive additions have bbeen
made. Figure 29.a shows ho ow the area at the bottom right changes to show details oof a
given cell of the histogramm as the user floats their mouse over it. Figure 29.b shoows
how selecting a particular fruit
f makes the histogram bars drop so it is possible to see
trends in the chosen fruit. We
W can now answer the question; in fact banana sales hhave
a slight, but steady increasin
ng trend.

Fig. 29. Interactive Stacked Hiistogram [30] (a) counts for individual cells (b) changing the bbase

You can probably think k of other simple interactive modifications to a chart llike
this, perhaps allow the useer to re-order the fruits by dragging the labels in the kkey,
maybe completely swoppin ng the visualisation to show side-by-side bars.
There are various differrent kinds of interaction, most of which we have seenn in
previous examples:
highlighting and focus – In the Cone Tree in figure 15, the user has selected one
of the nodes causing it and its parents to be highlighted and also rotated to the fronnt of
the view. This highlightin ng is useful even for simple elements, but more so for
examples such as this, wh hen the focus is not just a single point, but in some w way
spread across the visualisaation, as in this case with the node and its parents,, or
highlighting a line in a charrt such as the Paris–Lyon timetable in Figure 10.
drill down and hyperlin nks – In the TreeMap of images in Figure 17, the imagee of
the cat has been selected and
a expanded so that we can see more details. Sometim mes
items of interest are expand ded in place, as in this case, or an outliner or file choooser
when a folder is opened. Alternatively drilling into an element may open a frresh
window or page, so being more
m like a hyperlink.
overview and context – As well as seeing details it is often important to gett an
overview of all the data. Th he TreeMaps in Figure 17 do this showing an entire phhoto
f system (on the right). When we are seeing details, we
collection (on the left) or file
may have to hide or reduce this overview, but having some idea of the context of w what
we are looking at is importaant. This is sometimes achieved by having a separate hiigh-
level view, for example, wh hen you zoom in to a part of a picture, image editors offten
show a thumbnail of the wholew image with the currently selected portion markked.
Another random document with
no related content on Scribd:
occasional uncaused and arbitrary deviation from its path, as a
means of bringing atoms into collision and combination. Thus with
them “freedom of indifference” was the result of physical difficulties.
In the Christian Church the doctrine seems to have owed its wide
—though not universal—acceptance to equally non-ethical difficulties
of a theological kind. If God “foreknew from all eternity” the
transgression of Adam and all its consequences, how could it be
compatible with His justice to punish Adam and all his posterity for
faults foreseen by Adam’s Creator?[201] The difficulty of reconciling
the divine omniscience with the divine justice was supposed to be
avoided—in truth, it was only evaded[202]—by assuming that man
was created with a “free will of indifference,” so that obedience would
have been just as easy as transgression if man had chosen to obey.
In our own time the problem has assumed a rather different
complexion, owing to the enormous developments of mechanical
physical science, which began with Galileo and Descartes. Rigid
causal determination being assumed as a first principle of physical
science, the question arose whether the assumption should not also
be extended to the psychical sphere. If so extended, it seemed to
strike at the roots of moral responsibility, by making all human acts
the inevitable “consequences of circumstances over which we have
no control”; if not admitted, the rejection of the principle of rigid
causal determination has often been thought to amount to the denial
that there is any principle of rational connection in the psychical
sphere. Hence, while persons specially interested in the facts of the
moral life have frequently inclined to the more or less radical denial
of rational connection between the events of the psychical series,
others, whose special interests have lain in the direction of the
unification of knowledge, have still more commonly thought it
necessary to hold that human action is determined by antecedents in
the same sense and to the same degree as the occurrences of the
purely physical order.
It will be our object to show that these rival doctrines of
Indeterminism and Determinism, or Necessitarianism, are alike
irrational, alike incompatible with what in practice we understand as
moral freedom of action, and alike based upon the false assumption
that rigid mechanical determination is itself an actual fact, and not a
mere postulate of the special physical sciences, valid only so far as it
is useful. But before we enter upon our task, it is necessary to begin
with a statement as to the real meaning of ethical freedom itself.
Until we know what we mean by the kind of freedom we, as moral
beings, desire and think we ought to have, it will be useless to ask
whether we are or are not free.
§ 2. “Free” and “freedom” are manifestly what are called by the
logicians “privative” terms; they denote the absence of certain
restrictions. To be “free,” in whatever special sense you may use the
word, means to be free from something. What, then, are the typical
limitations which, in practice, we resent as making us unfree? They
seem to be, in the main, the following:—(1) We are not free when
our limbs are actually set in motion by an external physical agency,
human or non-human. And the reason why we are then unfree is that
the resulting movements of our bodies do not express a purpose of
our own. They either express the purpose of some other being who
moves our limbs as seems good to him, or, as in the case where we
are set in motion by the “forces” of the inanimate world, express no
purpose at all that is recognisable to us as such. And in either case
we have expressed no purpose of our own by our movements; they
do not truly belong to us at all, and there is therefore no freedom. It
is not necessary that the result of the movement should be one
which, if it had been suggested, we should have declined to entertain
as a purpose of our own. We might perhaps, if left to ourselves, have
done just what another man or the system of physical forces has
done for us. Still, so long as the deed, whatever it was, was done for
us and not by us, so long as it corresponded to no actual purpose of
ours, it was not a free act.
(2) Again, we are not truly free when we act in ignorance (not due
to previous free action of our own)[203] of the special circumstances.
Here there is, as there was not in the former case, a genuine act. We
actually purpose to do something, but what we purpose to do is not
the deed which results from our movements. E.g., if I shoot a
comrade by mistake for one of the enemy, it is true that I purpose to
shoot, and so far the shooting is an act, and a free act, of my own.
But I did not purpose to shoot my comrade, and so the result, in its
concreteness, is not the expression of my purpose, and I
consequently regard myself as not fully free in doing it, and therefore
not morally accountable for it. So far our analysis coincides with that
of Aristotle, previously referred to.
(3) Again, I am not acting freely where the circumstances are not
such as to admit of the formation of purpose at all. For this reason,
merely automatic action—if there is such a thing—is not genuine
action, and therefore not free.[204] Impulsive action without reflection,
again, comes under this category. It is, of course, accompanied by
feelings of satisfaction, and if impeded gives rise to craving, and so
cannot be called simply non-purposive. But in genuinely impulsive
reaction, where the possibility of reflection is excluded, there can be
little clear awareness of the concrete character of the purpose that is
being put into execution, and hence such action is not truly free. And
in practical life, though we are certainly held morally responsible for
impulsive action, in so far as it is thought we might have modified it
by previous habitual practice of reflection or by avoiding a situation
which we had reason to think would deprive us of the power to
reflect, we are never held as fully accountable for the deed of
impulse as for the reflectively thought out and deliberately adopted
purpose.[205]
Further, we feel ourselves unfree when we fail to execute our
purposes, either from sheer inability to attend to a consistent scheme
of action, or because we attend equally to purposes which are
internally incompatible. This is why the “democratic” man, whose
interests are an incoherent medley without logical unity, and the
“tyrannical man,” or, as we should now say, the “criminal type,”
whose passions are constantly at war with one another and with his
judgment, are regarded by Plato as the typically unfree beings. To be
really free, in the last resort, we must have purposes which are
coherent and abiding. And it is thus no paradox to say that
unfreedom in the end means, in the main, not knowing your own
mind, while to be free is to know what you mean.
§ 3. We may now draw some important consequences from this
review of the facts upon which every valid interpretation of freedom
has to be based. (1) Freedom, as Locke said in that famous chapter
“On Power” which is still the classic discussion of the whole subject
as far as English philosophy is concerned, “belongs to the man, not
to the will.” The proper question to ask is, “Am I free?” not “Is my will
free?” or “Have I a free will?” For “freedom” and “will,” as the facts
enumerated above show, are but the negative and the positive name
for the same property, the property of acting so as to put what we
first possessed as our private purpose into execution in the world of
sensible fact. I “will” when my outward deed is thus the expression of
my purpose; in the same case, and in no other, I am “free.” Thus to
“will” and to be “free” are one and the same thing; a will which was
not free would be a will which was not the translation into sensible
fact of any one’s purpose, and thus no will at all. Thus the question,
“Are we free?” might be also put in the equivalent form, “Can we
ever will anything?” and to the question, as thus put, experience
gives a ready answer. For we certainly do conceive purposes, and
we certainly, in some of our movements, do translate those purposes
in act. And therefore we may say that freedom is undoubtedly, in the
only sense in which it is desired, a fact of immediate experience.[206]
(2) If we retain the expression “freedom to will” by the side of the
phrase “freedom to act,” it can only be in a very special sense. It is
clear that not only may my outward deed be a translation into fact of
my present purpose, but my present purpose itself, as a psychical
event, may also be a translation into fact of a former purpose. This is
largely the case with all results of deliberate self-training and
discipline, and to a less degree with all acquired habits. Thus, e.g.,
the movements by which I write these lines are the expression of my
preconceived purpose to write the present paragraph, but that
purpose itself, as an event in my history, is similarly the expression
of a former purpose to compose a work on Metaphysics. Thus there
is a real sense in which we can agree with Leibnitz in criticising
Locke’s dictum that we are free to act, but not free to will. For the
mental conception of a purpose is itself an act, and in so far as it
translates into existing thoughts and feelings a previous purpose it
may be said itself to be “freely willed.”[207]
(3) Freedom, in actual experience, is always limited, and,
moreover, admits of the most various degrees. As to the first point, it
follows immediately from our consideration of the circumstances
which make us unfree. If to be fully free means that your outward
deed is the full expression of an inward consistent purpose, then we
can see at once that complete freedom is, for all finite beings, an
infinitely distant ideal. For it means (a) that I am not hampered in the
execution of my purpose by vacillation of interest or conflict of
incompatible interests within myself; (b) nor by the establishment of
“habitual” reactions so nearly mechanical as to repeat themselves
out of season unless checked by special reflection; (c) nor by the
limits set to my power to “act or to forbear” in the physical world by
the action of my fellows and of “brute” nature.[208] Hence only an
experience which is absolutely devoid of internal conflict and
external, partly discrepant environment, in other words, only the
experience which is the infinite whole, can be in all its detail entirely
and absolutely free. From the possibilities of internal lack of unity of
purpose and external collision with rival purpose which are
inseparable from our position as finite beings, it must follow that we
are never more than partially or relatively free.
And that the degree to which we are free varies with the nature of
our purposes and their relation to the environment, is also manifest.
There is an indefinite plurality of such degrees, ranging up from the
total or all but total absence of freedom in the case of directly
constrained motion up to the case of cordial co-operation with the
other members of a relatively self-supporting social group in the
conscious and systematic execution of an elaborate and coherent
scheme of action. To indicate the principal distinctions among such
grades of freedom which are of practical importance for law and
morality is the task of systematic Ethics, and need not be attempted
by us here. We may add that our investigation has made it apparent
that true moral freedom, of whatever degree, is no inalienable
heritage into which men step by the “accident of birth,” but—in the
main and as an actual possession—a prize which has to be won by
the double discipline of self-knowledge and self-mastery, and of
social comradeship, and may be, and is, forfeited by the neglect of
the arts by which it was first gained. No doubt one man’s inherited
disposition may make the practice of self-control, or again of social
fellowship, easier to him than to another, and to this extent we may
say that we are born with a greater or lesser “capacity for freedom,”
but of its actual possession we have all to say, “with a great price
purchased I this freedom.”
 (4) Finally, our examination of the facts of morality enables us to
define true freedom. We are free, as we have seen, just so far as our
experience is the embodiment of coherent and permanent interest or
purpose, and freedom is, like “will,” simply an abstract expression for
the teleological unity which, in varying degrees, is an essential
feature of all experience. Hence we can at once see that freedom
does not mean “absence of rational connection” or “absence of
determination,” but does mean, as so many recent philosophers
have told us, for us finite beings, self-determination. I am most free
when acting for the realisation of a coherent rational purpose, not
because my conduct is “undetermined”; in other words, because
there is “no telling” what I shall do next, but because it is, at such
times, most fully determined teleologically by the character of my
inner purposes or interests,—in other words, by the constitution of
my self. The more abiding and logically coherent my various
purposes in action, the freer I am, because it is my whole self or
system of rationally connected interests, and not the insistence of
others, or some passing whim or impulse which I may forthwith
disown as no part of my “true self,” which is getting expression in my
outward deeds. And if it were possible for a finite being to become
absolutely free, as we have seen that it is not, such a being would, in
the very moment of its entire deliverance, become also absolutely
determined from within; its whole life, as manifested to the outsider
in the series of its deeds, would become the perfect and systematic
expression of a single scheme of coherent purposes.
§ 4. We see, then, that such a genuine but limited freedom as is
really implied in the existence of morality is not only compatible with,
but actually demanded by, the principles of a sound Metaphysics.
From the side of morality we meet with the demand that human
beings shall be, in part at least, creatures whose outward acts shall
be the genuine expression of individual purpose; from the side of
Metaphysics we have already learned that just this teleological unity,
genuine though imperfect, is the essential nature of every finite
experience. We are now to see how a problem in itself quite simple
leads to insoluble difficulties and to the rival absurdities of
Indeterminism and Determinism when it is perverted by an initial
metaphysical blunder. The initial mistake of both the rival theories
consists simply in taking rigid mechanical determination of events by
their antecedents in accord with the principle of Causality as an
actual fact, the divergence between them only concerning the extent
of the sphere of existence for which such determination prevails.
According to the indeterminist, the action of conscious beings forms
a solitary exception to a principle of determination which is
absolutely valid for all purely physical processes. According to the
determinist, there are no exceptions to the principle, and our
confessed inability to predict the course of an individual life or a
period of history from general laws in the same way in which we
predict an eclipse or a display of leonids, is due merely to the greater
complexity of the necessary data, and the temporary imperfections
of our mathematical methods.
It should be noted that there is no substantial disagreement
between the more sober representatives of the two views as to the
actual facts of life. The indeterminist usually admits that in practice,
when you know enough of a man’s character and of the influences
brought to bear upon him, you can tell with some confidence how he
will conduct himself, and that social intercourse, education, and
penal legislation would be impossible if you could not. Similarly, the
determinist admits that it would be very rash to treat your predictions
of human behaviour in practice with absolute confidence, and that
the unexpected does frequently happen in human life. The dispute is
solely about the philosophical interpretation of facts as to which there
is virtually universal agreement. According to the determinist
interpretation, if you were put in possession of the knowledge of a
man’s “character” and of his “circumstances” (and it is assumed that
it is theoretically possible to have this knowledge), and had sufficient
skill to grapple with the mathematical problems involved, you could
calculate his whole behaviour in advance, from the cradle to the
grave, with infallible precision. According to the indeterminist, you
could not do so, and your failure would arise not from any theoretical
impossibility of obtaining the supposed data, but from their
insufficiency. Our behaviour, he alleges, is not exclusively
determined by the interaction of “character” and circumstances; even
with the complete knowledge of both these elements, human action
is incalculable, because of our possession of a “free will of
indifference” or power to act indifferently according to or in violation
of our “character.” You can never say beforehand what a man will do,
because of this capacity for acting, under any conditions, with equal
facility in either of two alternative ways.
I propose to show briefly that the determinist is right in saying that
conduct is completely determined by “character”—if the term be
understood widely enough—and circumstances, but wrong in holding
that this makes infallible prediction possible; on the other hand, that
the indeterminist is right in denying the possibility of such prediction,
but wrong in the reason he gives for his denial. Infallible prediction is
impossible, not because of the existence of “free will of indifference,”
but because the assumed data of the prediction are such that you
could not possibly have them until after the event. Finally, it will be
pointed out that the two errors both arise from the same false
metaphysical theory that the causal principle is a statement of real
fact.[209]
§ 5. Determinism. To begin with the view of the determinist.
Human conduct, he says, must be, like other processes,
unequivocally determined by antecedents, and these antecedents
must consist of (a) character and (b) external circumstance. For (1)
to deny the causal determination of our acts by antecedents is to
deny the presence of rational connection in the psychical sphere,
and thus to pronounce not only Psychology, but all the sciences
which take psychical events as their material and attempt to discover
rational connections between them, in principle impossible. Thus the
very existence of Psychology, Ethics, and History proves the
applicability of the principle of causal determinism to “mental states.”
(2) This is still more evident if we reflect that all science consists in
the formulation of “laws” or “uniformities,” and that the formulation of
“laws” rests upon the principle that “same result follows under same
conditions”—i.e. upon the principle of causal determination.
(3) Further, if psychical events are not so determined, then
Psychology and the mental sciences generally are inconsistent with
the general principles of the mechanical physical sciences.
(4) And, as a matter of fact, we do all assume that psychical
events are causally determined by their antecedents. In Psychology
we assume that our choices are determined by the strength of the
motives between which we choose. Hence, if you know what are the
“motives” present to a man’s choice, and the relative strength of
each, the determinist thinks the prediction of his conduct is reduced
to the purely mathematical problem of the solution of an equation or
set of equations. That our present mathematical resources will not
avail for the unequivocal solution of such equations is, on this view, a
mere temporary defect incidental to the present condition of
mathematical science. In principle the equations must be soluble, or
“there is no science of human action.”
(5) And in practical life we do all assume that it is possible to
predict with considerable confidence the effect of typical conditions
upon the aggregate of mankind, and also, when you have the
requisite data, the effect of a definite set of conditions upon an
individual man. Thus we count upon the deterrent effects of
punishment, the persuasive influence of advertisement, etc.; and
again, in proportion as we really know our friends, we believe
ourselves able to answer for their conduct in situations which have
not as yet arisen. Why, then, should we suppose it theoretically
impossible, if adequate data were furnished, to calculate the whole
career of a man or a society in advance, as the astronomer
calculates the path of a planet from its elements? These are, I think,
the chief of the stock arguments by which Determinism has been
defended. (With the purely theological argument from the
absoluteness of the divine foreknowledge I have already dealt in
passing, and do not propose to refer to it again.)
§ 6. It is not difficult to see that the logical value of all these
arguments is nothing at all. They fall of themselves into two groups,
one based upon the general view that all rational connection, or at
least all such rational connection as is significant for our knowledge,
is mechanical causal sequence, the other upon an appeal to the
supposed actual practice of the mental sciences. We may deal with
the first group (arguments 1 to 3) first. It is certainly not true that
causal determination by antecedents is the only form of rational
connection. For there is manifestly another type of connection, which
we have already seen to be fundamental for the mental sciences,
namely, teleological coherence. And we have learned in our
preceding books that no truly teleological or purposive series can
really be mechanically determined by uniform causal laws of
sequence, though it is often convenient for special purposes, as in
the physical sciences, to treat such a series as if it were
mechanically determined. Whether this type of procedure will be
valid in the mental sciences, depends upon the further question
whether our interest in the study of mental processes is of the kind
which would be satisfied by the formulation of a number of abstract
uniformities or laws of sequence, and the neglect of all those
features of real mental life of which such laws take no account.
In the physical sciences, as we saw, this mechanical scheme was
valid only because we have an interest—that of devising general
rules for dealing with typical physical situations—which is met by
neglecting all those aspects of concrete fact which the mechanical
scheme excludes. But we also saw that the nature of our interest in
psychological investigation was predominantly (and, in the case of
the study of voluntary action, exclusively) of a different kind. Our
interest in these investigations was to obtain such a teleological
representation of psychical processes as might be made available
for the appreciative judgments of Ethics and History and their
kindred studies. Thus, even admitting the possibility of treating
psychical life for some purposes, by abstraction from its teleological
character, as if it were a mechanical sequence, the abstraction would
be fatal for the purposes of the concrete mental sciences, and is
therefore inadmissible in them. A teleological unity in which we are
interested as a teleological unity cannot, without the stultification of
our whole scientific procedure, be treated in abstraction from its
teleological character.
This rejoinder to the first of the determinist’s arguments is at the
same time a refutation of the second. It is true that any science
which aims exclusively at the discovery of “laws” or “uniformities”
must adopt the causal principle, and must resolutely shut its eyes to
all aspects of concrete fact which cannot be resolved into
mechanical sequence of “same result” on “same conditions.” But, as
we saw in the first chapter of this book, the characteristic task of
Psychology, except in those parts of it which appear to be mere
temporary substitutes for the Physiology of the future, is not the
discovery of “laws of mental process,” but the representation in
abstract and general form of the teleological unity of processes
which are the expression of subjective interests. Psychology, then, in
its most characteristic parts, is not based upon the causal postulate
of mechanical science, but on the conception of teleological
continuity.
Our answer to the determinist’s third argument is therefore that we
admit the truth of the allegation that Psychology and all the more
concrete mental sciences which make use of the symbolism of
Psychology, because essentially teleological in their view of mental
process, would be inconsistent with the mechanical postulates, if
those postulates had any claim to admission into mental science as
its ruling principles. We deny, however, that they have any such
claim to recognition. Being, as we now know that they are, mere
methodological rules for the elimination from our data of everything
which is teleological, the mechanical postulates are only legitimate in
Psychology so far as Psychology desires mechanical results. How
far that is, we have learned in the first two chapters of the present
Book, and we have found that the initiation of purposive action is not
a process which Psychology can fruitfully treat as mechanical.
§ 7. Turning now to the determinist’s allegations as to the factual
procedure of the mental sciences, we may make the following
observations:—(1) As to the argument from the psychological
treatment of “motives” as the determining antecedents of choice, we
say that it is either an empty tautology or a fallacy, according to the
sense you please to put on the much-abused term “motive.” Choice
is causally determined by the “strongest motive”; what does this
mean? If the “strongest motive” simply means the line of action we
do in fact choose, the argument amounts to the true but irrelevant
observation that we choose what we do choose, and not something
else. But if “motives” are to be regarded as antecedents causally
determining choice in proportion to their strength, as mechanical
“forces” determine the path of a particle in abstract Mechanics, we
must suppose the “strength” of the various “motives,” like the mass
of an attracting body, to be previously fixed, independent of the
choice they determine. In other words, the determinist argument
requires us to hold that alternative possibilities of action are already
“motives” apart from their relation to the purpose of the agent who
has to choose between them, and moreover have, also in
independence of the purpose or “character” of the chooser, a
“strength” which is in some unintelligible way a function of—it would
not be easy to say of what, though it is incumbent on the determinist
to know. And this seems no better than rank nonsense. An
alternative is not a “motive” at all, except in relation to the already
existing, but not fully defined, purpose of some agent, and whether it
is a “strong” or a “weak” motive depends likewise on the character of
the agent’s purpose. The attempt to conceive of “motives” as
somehow acting on a mind with an inherent “strength” of their own,
as material particles attract other material particles proportionately to
their masses, is so palpable an absurdity, that nothing more than the
candid statement of it is needed for its complete exposure.
And (2) there is an equal absurdity inherent in the determinist view
as to the kind of prediction of conduct which is possible in concrete
cases. We have seen already in our Third Book that no infallible
prediction of the course of events in an individual case is ever
possible. Mechanical calculation and prediction we found to be
possible in the physical sciences simply because they deal with the
average character of a vast aggregate of processes which they
never attempt to follow in their concrete individual detail. And
trustworthy prediction of human conduct by the aid of “causal laws”
was seen to be of the same kind. Your uniformities might hold good,
so long as they professed to be nothing more than statistical
averages got by neglecting the individual peculiarities of the special
cases composing them, but nothing but acquaintance with individual
character and purpose would justify you in making confident
predictions as to the behaviour of an individual man.
Now, when the determinist says, “if you knew a man’s character
and his circumstances you could predict his conduct with certainty,” it
is not this kind of individual acquaintance which he has in view. He
means that the “character” of an individual man could be reduced to
a number of general formulæ or “laws of mental action,” and that
from these “laws,” by simply putting them together, you could
logically deduce the man’s behaviour. To see how irrational this
assumption is, we need only ask what is meant exactly by the
“character” which we suppose given as one of the elements for our
supposed calculation. If it means the sum-total of the congenital
“dispositions” with which we are born, then—apart from the difficulty
of saying precisely what you mean by such a “disposition”—the
determinist statement is not even approximately true. For (a) though
it may be true that a man’s behaviour in a given situation is an
expression of his “character,” yet the “character” is not the same
thing as “congenital disposition.” Disposition is the mere raw material
of the “character,” which is formed out of it by the influence of
circumstance, the educational activity of our social circle, and
deliberate self-discipline on our own part. And the “character” thus
formed is not a fixed and unvarying quantity, given once and for all at
some period in the individual’s development, and thenceforward
constant; it is itself, theoretically at least, “in the making” throughout
life, and though you may, from personal intimate acquaintance with
an individual man, feel strongly convinced that his “character” is not
likely to undergo serious changes after a certain time of life, this
conviction can never amount to more than what we properly call
“moral” certainty, and is never justified except on the strength of
individual familiarity.
(b) This leads us to our second point. If—to suppose the
practically impossible—you did know a man’s “character” with the
knowledge of omniscience, you would clearly also know every act of
his life. For his “character” is nothing but the system of purposes and
interests to which his outward deeds give expression, and thus to
know it completely would be to know them completely too. But—and
this is what the determinist regularly overlooks—you could not
possibly have this knowledge of the man’s “character” until you were
already acquainted with the whole of his life. You could not possibly
thus know “character” as a datum given in advance, from which to
calculate, with mathematical precision, the as yet unknown future
acts of the man in question, because, as we have seen, the
“character” is, in fact, not there as a given fact before the acts
through which it is formed. Your data could at best be no more than
a number of “dispositions” or “tendencies,” and from such data there
can be no infallible prediction, because, in the first place,
“dispositions” are not always developed into actual fixed habits; and,
in the second, your data, such as they are, are incomplete, seeing
that “dispositions” may, and often do, remain latent and escape
detection until the emergence of a situation adapted to call them out.
So that, even if it were true that complete knowledge of a man’s
original stock of “dispositions” would enable you to calculate his
career from its elements, it would still be impossible to be sure that
your knowledge of his “dispositions” was complete.
Thus, if a “science of human nature” really means a power to
calculate human conduct in advance from its elements, we must
admit that there is not and can be no such science. As a fact,
however, what we really mean by a “science of human nature,” when
we speak of it as possible or as partly existent already, is something
quite different. We mean either Psychology, individual and social,
which is simply an abstract symbolism for the representation of
teleological process in its general nature, or History, which is the
detection of coherent purpose in human action, after the event; or,
again, Ethics and Politics, which are appreciations of such purpose
by an ideal standard of worth. Not one of these sciences has ever
attempted the calculation of human action in advance by general
laws; such forecasts of the future as we do make, with rational
confidence, are palpably based, wherever they are of value, on
concrete experience, our own or that of others, and not upon the
principles of an imaginary mechanics of the human mind.
§ 8. Indeterminism. With the fallacies of the indeterminist we must
now deal more briefly. This is the more possible as Indeterminism,
though common enough in popular moralising, has never won
anything like the position of the rival doctrine as the professed creed
of scientific investigators. The essence of the indeterminist position
is the denial of the principle affirmed alike by the doctrine of self-
determination and, in an unintelligent travesty, by the determinist
theory that conduct results from the reaction of “character” upon
circumstances. Seeing that, if all human action is mechanically
determined in advance by its “antecedents,” and is thus theoretically
capable of being deduced from its “elements,” there can be no true
moral freedom, and, not seeing that the essence of true freedom is
teleological as opposed to mechanical determination, the
indeterminist thinks himself compelled to assert that human action is,
in the last resort, not “determined” even by human character. There
is a “free will of indifference” inherent in human nature, in virtue of
which a man’s acts, or at least those of them in respect of which he
is morally “accountable,” are free, in the sense of being independent
of his character.
Freedom, according to this view, consists in the ability indifferently
to adopt either of two alternative courses; so long as one alternative
is closed to you (whether by your “character” or by external
circumstances makes no difference according to the indeterminist),
you are not “free” and not acting as a moral and accountable being.
You are only acting freely in following your purpose when you could
equally well follow its direct opposite. The arguments by which this
doctrine is supported, over and above the general contention that
determination by antecedents is incompatible with moral
responsibility, are chiefly of the nature of appeals to immediate
feeling. Thus we are told (1) that when we act from choice and not
under compulsion we always have the immediate feeling that we
could equally well act in the opposite sense; and (2) that it is a
matter of direct experience that, in resisting temptation, we can and
do act “in the line of greatest resistance,” and that the “will” is
therefore independent of determination by “motives.”
The detailed discussion of the actuality of the alleged facts
belongs, of course, to Psychology, and I do not propose to enter into
it here. But it should be manifest that, even admitting the facts to be
as the indeterminist states them, they do not warrant the inference
he bases on them. Thus (1) it is no doubt true that I often am aware,
in resolving on a certain course of action, that I could, if I pleased,
act differently. But the conditional clause by its presence makes all
the difference between teleological determination and no
determination at all. It is, e.g., no genuine fact of experience that I
am aware that I could violate all the habits of a lifetime, practise all
the crimes I most abhor, and neglect all the interests to which I am
most devoted. I could do all this “if I pleased,” but before I could
“please” I should have to become a different man; while I am the
man I am, it is a manifest absurdity to hold that I can indifferently
express in my behaviour the purposes which constitute my
individuality or their opposites.
 (2) The argument from the successful resistance of temptations is
equally fallacious. We have seen already that the determinist
assumption against which it is directed, namely, that conduct is
mechanically determined by the inherent “strength” of “motives,” is
itself unmeaning. “Motives” are, if they are anything, another name
for the interests which constitute our character, not external
influences which “work upon” that character, and thus their relative
“strength” is nothing independent of character, but a new expression
for the structure of the individual character itself. But the counter-
argument of Indeterminism is just as unmeaning. To talk of the
“conquest” of temptation as the “line of greatest resistance” is to use
the very same unintelligible mechanical analogy as the determinist
uses in talking of the antecedent “strength” of a “motive.” There are,
in fact, only two possible interpretations of the indeterminist’s
contention, and neither of them supports his conclusion. Either the
“resistance” of which he speaks must be measured by our actual
success in resisting the suggestion to act, and in that case the very
fact that we do not yield to the temptation shows that for us yielding
would have been the “line of greatest resistance”; or else
“resistance” must be measured by the extent to which the rejected
alternative still persists as a psychical fact after its rejection. Then
the alleged experience simply amounts to this, that we can and
sometimes do, in obedience to training or conviction, refuse to act
upon suggestions which as psychical facts have sufficient intensity to
remain before the mind even after our refusal. And this, interesting
and suggestive as it is, seems no particular reason for denying the
teleological determination of our conduct.[210]
The real metaphysical objection to Indeterminism however, is not
that it is an unprovable and unnecessary hypothesis, but that it
involves the denial of rational connection between human actions.
By declaring that conduct is not determined by character, it virtually
asserts that it is chance which ultimately decides how we shall
actually behave in a concrete case. And chance is simply another
name for the absence of rational connection. This is illustrated, e.g.,
by the use we make of the conception of chance in the various
empirical sciences. Thus, when I say that it is a matter of chance
what card I shall draw from the pack, what I mean is that the result
depends in part upon conditions which I do not know, and therefore
cannot use as data for a conclusion in favour of one result rather
than another. I do not, of course, mean that the result is not
conditioned at all, or that, with a sufficient knowledge of the
conditions it might not have been calculated in advance, but merely
that I in particular have not this sufficient knowledge. Hence the
admission of chance in the relative sense of “conditions not at
present accurately known” does not conflict with the fundamental
axiom of all thinking, the principle that all existence is a rational unity
or scheme of some sort. In fact, since we never can know the
“totality of the conditions” of anything, it would be true to say that
there is an element of chance, in this relative sense, in all concrete
actualities.
But absolute chance, such as the doctrine of an indeterminate free
will maintains, would amount to the simple absence of any rational
connection whatever between the facts which are alleged to issue
from such a will. This is why the indeterminist view leads in the end,
if consistently carried out, to the same metaphysical absurdity as the
determinist. From failure to see that rational connection, such as is
presupposed when we impute praise or blame to an agent on the
score of his conduct, means teleological determination, both the rival
theories in the end deny the rational interconnection of human acts,
the one replacing it by the fiction of a purposeless mechanical
“necessity,” the other by the equal fiction of a “blind chance.” And the
two fictions are really the same thing under different names. For the
only piece of definite information that could be extracted either from
the assertion that human conduct is mechanically determined, or that
it is the result of chance, is the conclusion that in either case it is not
the expression of coherent purpose.
§ 9. It is thus obvious that Indeterminism fails, in precisely the
same way as the opposing theory, to afford any theoretical basis for
moral responsibility. True, I cannot be “responsible” for deeds which
are the outcome of a purely mechanical system of antecedents,
because such deeds, not issuing from the purposes of my self, are in
no true sense mine; but the same would be equally true of the
results of an indeterminate free will. As not owing their existence to
my purpose, those results are in no real sense “my” acts, and the
choice of the name “free will” for their unknown source only serves to
disguise this consequence without removing it. Only as issuing from
my character, and as the expression of my individual interests, can
acts be ascribed to me as “mine” and made the basis of moral
approbation in censure of my “self.”
Thus we see that the determinist and the indeterminist are led
alike to impossible results because of the common error involved in
their point of departure. Both start with the false assumption that the
causal determination of an event by its “antecedents”—which we
have in our earlier books seen to be a postulate ultimately not in
accord with reality, but permissible in so far as it permits us to obtain
useful results by treating events as if they were thus determined—is
ultimately real as a feature of concrete existence. Having thus at the
outset excluded genuine teleological determination from their
conception of the world of change, both theorists are alike debarred
from the correct understanding of those psychical processes for the
comprehension of which teleological categories are indispensable.
In the terms of theories which treat determination as purely
mechanical, the factors which manifestly are the determining
conditions of conduct, namely, character and the alternative
possibilities of action, inevitably come to be conceived of as the
temporal “antecedents” of the act which issues from them. And when
once this notion of character as a sort of pre-existing material upon
which “motives” from without operate has been framed, it matters
little in principle whether you take “character” and “motive” by
themselves as the complete antecedents by which action is
determined, or add a third “antecedent” in the form of an inexplicable
arbitrary “free will.” In either case all possibility of a truthful
representation of the freedom actually implied in moral accountability
was surrendered when the “character” which expresses itself
through an act, and the “motive” which is another name for that
character as particularised by reference to circumstances, were
falsely separated in thought from each other, and then further treated
as the temporal antecedents of the act in which they are expressed.
In our own treatment of the problem of freedom we were able to
escape both sides of the dilemma, because we recognised from the
first that the categories of mechanical determination are not the
expression of real fact, but limitations artificially imposed upon facts
for special purposes of a kind which have nothing in common with
the ethical and historical appreciation of human conduct, and
therefore irrelevant and misleading when applied out of their rightful
sphere.

Consult further:—H. Bergson, Sur les données immédiates de la

conscience; F. H. Bradley, Ethical Studies, Essay 1; W. R. B. Gibson,
“The Problem of Freedom” (in Personal Idealism); T. H. Green,
Prolegomena to Ethics, bk. i. chap. 3, bk. ii. chap. 1; W. James,
Principles of Psychology, vol. ii. chap. 26; Will to Believe (The
Dilemma of Determinism); J. Locke, Essay concerning Human
Understanding, bk. ii. chap. 21 (on Power); J. Martineau, Types of
Ethical Theory, vol. ii. bk. i. chap. 1; J. S. Mill, Logic, bk. vi. chap. 2
ff.; J. Royce, The World and the Individual, Second Series, lect. 8; H.
Sidgwick, Methods of Ethics, bk. i. chap. 5; Lectures on the Ethics of
Green, etc., pp. 15-29.

200. See Methods of Ethics, bk. i. chap. 4, § 6 (pp. 72-76 of 5th

ed.).
201. So Omar Khayyám—

“Oh Thou, who didst with pitfall and with gin

Beset the Road I was to wander in,
Thou wilt not with Predestined Evil round
Emmesh, and then impute my Fall to Sin.”
(Fitzgerald, ed. 4, stanza 80.)
And our own poet—

“Thou madest man in the garden; thou temptedst man, and he

fell,” etc.
(For the original of the stanzas on Predestination in Fitzgerald’s
Omar, see, e.g., the Persian text of Whinfield, quatrains 100, 126,
197.)
202. Evaded, because, even granting the satisfactoriness of the
solution for the special case of Adam, there would still be the
problem of reconciling the alleged “free will” of his descendants with
their inheritance of “original sin.” The more rigid Calvinism, with its
insistence on the natural corruption of man’s heart and the
absoluteness of predestination, seems to secure logical consistency
at the expense of outraging our moral convictions. Like so many
popular theological problems, this of the conflict between God’s
omniscience and justice arises from a misconception of the issue. It
is only when the category of time is illogically applied to the ex
hypothesi perfect, and therefore timeless, nature of God that God’s
knowledge comes to be thought by as fore knowledge before the
event, and thus occasions the difficulty which the “free-will” theory
was intended to remove. See on this point, Royce, The World and
the Individual, vol. ii. lect. 8, and compare Bradley, Ethical Studies, p.
19. Of course, the case would be altered if we thought of God as
finite and imperfect, and therefore in time. But there would then be
no longer any reason for believing either in His omniscience or His
omnipotence, and so no problem would arise.
203. Remember that abstention from acting is itself action, just as
in Logic every significant denial is really an assertion. Hence our
proviso meets the case of wilful neglect to inform myself of the
material circumstances.
204. The only automatic acts of which we really know the
psychical character are our own “secondarily automatic” or “habitual”
acts. It is, of course, a problem for the casuist how far any particular
reaction has become so completely automatic as to be no longer an
occasion for the imputation of merit or guilt.
205. For purposes of law it may often be impossible to draw the
distinction, and we may have to acquiesce in the rough-and-ready
alternative between entire accountability and complete non-
accountability. But in passing moral judgment on ourselves or others
in foro conscientiæ, we always recognise that accountability is a
thing of degrees. On this point see Mr. Bradley’s previously quoted
article in Mind for July 1902.
206. It must, however, be carefully noted that will in the sense in
which it is equivalent to freedom must be taken to include what some
writers, e.g., Bradley, call a “standing” will—i.e. any series of acts
originally initiated by an idea of the resultant changes, which is