Professional Documents
Culture Documents
Lecture 4: Requirements Elicitation Elicitation Techniques: University of Toronto University of Toronto
Lecture 4: Requirements Elicitation Elicitation Techniques: University of Toronto University of Toronto
University of Toronto Department of Computer Science University of Toronto Department of Computer Science
1
University of Toronto Department of Computer Science University of Toronto Department of Computer Science
Example of Interviews
hard data
Types:
Questions: Structured - agenda of fairly open questions
Open-ended - no pre-set agenda
What does this data tell
you? Advantages
What would you do with Rich collection of information
this data? Good for uncovering opinions, feelings, goals, as well as hard facts
Can probe in depth, & adapt followup questions to what the person tells you
Disadvantages
Large amount of qualitative data can be hard to analyze
Hard to compare different respondents
Interviewing is a difficult skill to master
Watch for
Unanswerable questions (“how do you tie your shoelaces?”)
Tacit knowledge (and post-hoc rationalizations)
Removal from context
Interviewer’s attitude may cause bias (e.g. variable attentiveness)
© 2000-2004, Steve Easterbrook 5 © 2000-2004, Steve Easterbrook 6
Source: Adapted from Goguen and Linde, 1993, p154.
University of Toronto Department of Computer Science University of Toronto Department of Computer Science
2
University of Toronto Department of Computer Science University of Toronto Department of Computer Science
University of Toronto Department of Computer Science University of Toronto Department of Computer Science
Disadvantages Notes:
May create unnatural groups (uncomfortable for participants)
Choose workshop participants carefully
Danger of Groupthink they should be the best people possible representing various stakeholder groups
May only provide superficial responses to technical questions Workshop should last 3-5 days.
Requires a highly trained facilitator Must turn a group of participants into a team - this takes 1-2 days.
Session leader makes sure each step has been completed thoroughly.
Watch for
Session leader steps in when there are differences of opinion - “open issues”.
Meeting room should be well-equipped for presentations, recording etc.
sample bias
dominance and submission
© 2000-2004, Steve Easterbrook 11 © 2000-2004, Steve Easterbrook 12
3
University of Toronto Department of Computer Science University of Toronto Department of Computer Science
rationalise
Background Observe behaviour of an agent as black box
Knowledge elicitation is concerned with It acts as if it has some knowledge about its Environment
discovering ‘expert’ knowledge environment which it uses rationally
Example
Example Techniques
Behaviour
Grew out of Expert Systems work in the Techniques It takes actions to achieve ascribed goals
Eliciting
Elicitingdomain
domainknowledge
knowledge observe
80’s Construct two models:
Card Sorting
Originally focussed on deriving expert’s Card Sorting Symbol Level - descriptions for mechanising Observer
Laddering
Laddering (modeller)
“rules” for Rule-based Systems Proximity Scaling Techniques behaviour
Proximity Scaling Techniques
More recently, focussed understanding Knowledge Level - descriptions of the agent's
mechanise
Eliciting
Elicitingperformance
performanceknowledge
knowledge
“problem solving methods” knowledge of the world
Protocol Analysis Agent
Protocol Analysis
Using
UsingMultiple
MultipleExperts
But KE is hard Experts
Delphi Technique Two-step rationality:
Delphi Technique
Separation of domain knowledge from Focus Groups Agent applies its knowledge in two stages: Symbol level
Focus Groups
performance knowledge Repertory Grids
Repertory Grids First creates a task specific model from the KL model
Modeling problems Automated
AutomatedTechniques
Techniques model based on features of the task.
Machine Learning
Brittleness Machine Learning Hence, we actually need 3 models: Knowledge
Assumption of rationality
Domain model - a systematic way of talking
Representational Problem about a domain, with a coherent ontology. Problem
epistemological inadequacy Task model - models goals, what it means to solving Behaviour
expressiveness vs. acquirability
achieve a goal, and how goals are related. method
Expert Bias Problem-solving method - a way of relating task
Task features
and domain models to accomplish goals.
© 2000-2004, Steve Easterbrook 13 © 2000-2004, Steve Easterbrook 14
University of Toronto Department of Computer Science University of Toronto Department of Computer Science
4
University of Toronto Department of Computer Science University of Toronto Department of Computer Science
University of Toronto Department of Computer Science University of Toronto Department of Computer Science
5
University of Toronto Department of Computer Science University of Toronto Department of Computer Science
post-modern view:
“there is no privileged viewpoint; all observation is value-laden; scientific
investigation is culturally embedded”
In RE: “validation is always subjective and contextualised”
Use stakeholder involvement so that they ‘own’ the requirements models
Use ethnographic techniques to understand the participant’s worldviews
© 2000-2004, Steve Easterbrook 23