Professional Documents
Culture Documents
Algorithmic governance is slowly being embraced by administrations all over India. As AI and
machine learning begin to play a prominent role in decision making, the use of new technologies
brings fresh challenges. The consequences of Algorithmic bias have proven to be unfair, dire and
discriminatory. This brief shall explore the underlying causes of the bias, the impact it has on
governance and the steps taken worldwide to combat the same. It also analyzes the jurisprudential
scenario in India and recommends some legislations through which the State of Karnataka can
implement Impact assessments and introduce Data Regulations to mitigate the risks associated
with Artificial Intelligence in governance.
Understanding Algorithmic Bias
While algorithms are used in many contexts, here the focus is on computer models that make
inferences from data about people, including their identities, their demographic attributes, their
preferences, and their likely future behaviors, as well as the objects related to them. The use of
artificial intelligence and machine learning systems introduces algorithmic bias. Algorithmic Bias
essentially refers to systematic and repeatable errors in a computer system that create unfair
outcomes. Bias can emerge due to the design of the algorithm and the unintended and
unanticipated decisions relating to the way data is coded, collected, selected or used to train the
algorithm.
Such biases have begun to plague administrations who have evolved modes of algorithmic
governance. Algorithmic governance is an alternative form of governance where the usage of
computer algorithms, especially of artificial intelligence, is applied to regulations, law enforcement,
and generally any aspect of everyday life such as transportation or land registration. Therefore the
emergence of new technological methods in administration have brought the need to have policy
frameworks that regulate transparency and accountability of the algorithmic tools used for
automated decision making.
A Global Perspective
There have been consequences of algorithmic bias around the world, some examples include :
1. The Netherlands, where the government employed an algorithmic system SyRI (Systeem
Risico Indicatie). It intended to detect citizens perceived as being at high risk for
committing welfare fraud, which quietly flagged thousands of people to investigators. This
caused a public protest. The district court of Hague shut down SyRI referencing Article 8 of
the European Convention on Human Rights (ECHR)1.
2. The United States of America,where multiple states implement predictive analytics as part of
their child protection system. Illinois and Los Angeles shut these algorithms down due to a
high rate of false positives2.
1
W
elfare surveillance system violates human rights, Dutch court rules
2
C an Big Data Help Save Abused Kids?
The growing number of such instances has lead to the development of legislations in Europe and
the United States aimed at reducing machines biases these include:
1. The “European Commission's Ethics Guidelines for Trustworthy3” AI prohibits solely
automated decisions in cases where there could be significant legal impact on the individual. A
right to human-in-the-loop and a non-binding right to explanation is also provided for.
2. “Algorithm Accountability Act4” was introduced in the US Congress in April 2019.It ensures
that companies will have to evaluate a broad range of algorithms — including anything that
affects consumers’ legal rights, attempts to predict and analyze their behavior, involves large
amounts of sensitive data, or “systematically monitors a large, publicly accessible physical
place.
3. “Algorithm Transparency Bill”5 was passed by the New York City Council in 2017. It was the
first of its kind in the United States, It established a task force to examine the hidden
algorithms governing life in New York and suggested a way for experts to study those tools for
error and bias.
The Current Indian Scenario
There is no clear regulatory framework in India to address bias in administration due to AI, as the
constitutional and administrative law is not geared towards assessing decisions made by
non-human actors. While the Personal Data protection Bill 2019 does provide provisions to access
and confirm personal data, it does not require computer model decisions to be explainable.
Similarly the SDPI(Securing Sensitive Personal Data or Information) rules fails to cover machine
biases as well.
Section 4 of the Right To Information Act 20056 requires public authorities to provide reasons for
its administrative or quasi-judicial decisions to affected persons. The problem arises when this
obligation remains unfulfilled. Algorithmic Bias can lead to public authorities being unable to
explain the outcomes of AI systems used in administration, particularly those of machine learning
systems which ‘learn’ from analysing patterns and data and follow logic that isn’t explicitly
programmed into it.
Moreover most of the AI systems are proprietary systems that are licenced for government use.
This brings about a clash between private and public interests and those most reliant on the state
for the delivery of public services like health and education or schemes for social security will be
disproportionately affected as such biases lead to lack of accountability and transparency.
Consequences of Bias in Emergencies
Contact tracing’ technologies are being extensively used in domains like disaster relief. Given the
extensive use of AI in emergency situations such as the COVD -19 pandemic the consequences of
3
Shaping Europe's digital future: Commission presents strategies for data and Artificial Intelligence
4
India needs to bring an algorithm transparency bill to combat bias
5
Id
6
Right to Information Act 2005
faulty decisions as a result of algorithmic bias are dire.The central government’s Aarogya Setu App7
allows users to self-assess their symptoms, and the application compartmentalizes them into
different groups based on their COVID-19 risk.If Algorithmic bias leads such systems to wrongly
urge people to pre-emptively take tests, then there is a risk that public health systems may be
overwhelmed prematurely. Moreover algorithmic bias also raises the possibility of false positives as
well as false negatives which could lead to unfortunate targeting of healthy people and the
dangerous exposure of COVID-19 carriers in society.
The creation of such systems also raises privacy concerns amongst people making it harder for the
government to receive public co -operation during emergencies. A solution to this would be the
introduction of interactive interfaces in apps that allow the users to decide what information they
desire to share.Thus increasing the chances of legitimate information input and reducing
algorithmic bias.
Dawn of Algorithmic Governance in Karnataka
Karnataka has been one of the first states in India to establish a policy on ‘Artificial Intelligence’
(AI) and machine learning technologies. In the Action Plan for the Development of Machine
Intelligence, the ‘Karnataka Jnana Ayog’ has advocated for the use of AI in governance in fields
including health, transport, education and e-governance8. The State government has stressed upon
the importance of artificial intelligence, data gathering and digitisation and how they assist in
making governance easy, transparent and efficient. It has already begun using AI in order to process
farm loan waivers and is using Facial Recognition for policing in Public areas9.
The most recent use of Artificial Intelligence in Karnataka was to battle the COVID-19 pandemic.
The government of Karnataka has launched a mobile application ‘Corona Watch’. The mobile
application aims to track the movement history of persons tested positive, before their detection in
order to take precautions and to contain the coronavirus outbreak.In addition to ‘Corona Watch’
app, the Karnataka government has also constituted a ten member Task Force on COVID-19. The
task force is utilizing IT applications & GPS to track individuals who are suspected of novel
CoronaVirus infection10.
Way Forward For Karnataka
The growth of algorithmic governance in Karnataka legitimizes the need for legislations that
combat algorithmic bias in the state. In order to ensure efficient and comprehensive
evidence-based justifications to the public, the operators need to be able to explain the AI outputs
and impact assessments, and reducing algorithmic bias will help them in the process. In order to
accomplish this, the State Government can establish an AI Alliance (AIA). The AIA will
comprise notable persons from the AI industry and academia based out Bengaluru with the
7
Arogya Setu wristbands to monitor patients movements
8
Karnataka ministers to soon get introduction in Artificial Intelligence
9
Id
10
How Karnataka govt uses technology, data and surveillance to contain Covid-19
Principal Secretary of IT,BT and S&T being the Member Secretary for the alliance.. The functions
of this alliance would be to
1. Assist and advise the government in drawing up a framework under the ambit of rules of the
‘Karnataka Transparency in Public Procurement Act, 1999’ Act to require an ‘algorithmic
impact assessment’ prior to any procurement of AI systems.
The Algorithmic Impact Assessment (AIA) framework proposed in this brief is designed to
support the government as it assesses the claims made about these systems, and to determine
where – or if – their use is acceptable, whether they are producing disparate impacts, and how to
hold them accountable. Section 1711 of the ‘Karnataka Transparency in Public Procurement
Act, 1999’ will need to be referred to.
2. Draw up standards to assess automated systems against risks of bias and opacity. This can
ensure that any proprietary system is made auditable and open to scrutiny to the government
3. Draw up procurement guidelines for the Government of Karnataka to enable departments/
agencies to evaluate and procure AI systems. This would also involve providing definitions for
the types of decision systems including but not limited to
a. "High-risk" automated decision systems; and
b. "High-risk" information systems.
4. Assist the government to make changes to the Right To Information Rules, 2005, framed
under Section 27 of the Right To Information Act, 2005 and include rules concerning the
transparency and explainability of automated decisions.
Such rules can possibly include -
a. Individuals have the right to know when government agencies have made decisions
concerning them which substantially involve the use of AI systems.
b. Define the broad parameters of such a decision, in a non-technical and simple
manner.
c. Incorporate a mechanism for individuals to seek a reasoned basis for the judgement
of such an automated system or opt for the decision to be reconsidered without the
use of such a system.
5. Develop an AI Action Plan for the state by developing on the groundwork laid down by
Karnataka Jnana Ayog in the ‘Action Plan for the Development of Machine Intelligence’.
Concluding Thoughts
In the absence of a regulatory framework, the expanding use of automated technologies can pose
significant concerns for the public, and can potentially violate constitutional norms of equality and
natural justice, the need of an AI regulations is the need of the hour for to protect both
government and private interests. A strong legal framework to combat Algorithmic Bias integrated
11
Power to obtain information- Notwithstanding anything contained in this Act or in any other law for the time
being in force, the Government may with a view to ensuring transparency call for and obtain, from any
Authority under this Act, any information relating to any matter in the process of procurement
with the potential of Algorithmic Governance can surely ensure a robust environment for the
growth of AI in Karnataka.