Topics In Demand
Notification
New

No notification found.

NASSCOM Feedback on TEC draft Framework for Fairness Assessment of AI/ML Systems
NASSCOM Feedback on TEC draft Framework for Fairness Assessment of AI/ML Systems

March 29, 2022

287

0


Listen to this article



In an effort to build more public trust in Artificial Intelligence (AI)/Machine Learning (ML) System, the Telecommunication Engineering Center (TEC), Department of Telecom has initiated a public consultation to develop a Framework for Fairness Assessment of AI/ML Systems.

As part of this, TEC had sought the stakeholders view on some specific questions. The questions covered the various aspects of fairness assessment of AI, like the appropriate classification of AI/ML systems for fairness assessment, type of bias in AI/ML systems, fairness metrics used by the industry and the procedures for handling code and training datasets.

In our feedback, we had noted that there are certain basic challenges in developing a Framework for Fairness Assessment of AI/ML Systems. The key challenges are listed below.

  • Fairness is a normative concept. There is no one universally agreed upon, or universally applicable, definition of “fairness” that can be tied to AI/ML systems. [1] For example, definitions may focus on formal conceptions of fairness that are based on equality (giving each individual or group the same resources) and not necessarily on substantive conceptions of fairness that are based on equity (giving individuals or groups the resources they need to succeed).[2] There are scenarios where equity, not equality is the meaningful outcome to be desired, but the definition of “fairness” being used cannot accommodate this.
  • Studies have shown that fairness perceptions do vary across cultures and efforts to assess the fairness of AI/ML systems in the Indian context must first arrive at an understanding of “fairness” contextualised to the Indian context.[3] At present, in India, while the NITI Aayog has published an approach paper with a set of guiding principles for AI/ML systems[4], neither they nor any other body of the Government of India has adopted an official definition of fairness.
  • There are inherent difficulties in putting abstract concepts into practice. Any initiative around standard setting in these aspects, needs to ensure that it can be operationalised. To that extent, all technological and methodological challenges need to be considered while framing the standard.
  • The concept of “fairness” has not reached a stage of sufficient clarity where certain standards can be prescribed, even if it is meant to be adopted voluntarily.

While we pointed out these broad challenges, we have shared our input on the specific questions that have been raised in the consultation document. For more details on this, please refer to the attached submission.

 

[1] For an illustrative list of widely used definitions of fairness, see N. Mehrabi, A Survey on Bias and Fairness in Machine Learning, ACM Computing Surveys, 54(6), (2022), available at https://arxiv.org/pdf/1908.09635.pdf

[2] Id.  

[3] See N. Sambasivan et al, Reimagining Algorithmic Fairness in India and Beyond, (2021) available at https://arxiv.org/pdf/2101.09995.pdf

[4] NITI Aayog, Responsible AI – Approach Document for India, (2021), available at https://www.niti.gov.in/sites/default/files/2021-02/Responsible-AI-22022021.pdf


That the contents of third-party articles/blogs published here on the website, and the interpretation of all information in the article/blogs such as data, maps, numbers, opinions etc. displayed in the article/blogs and views or the opinions expressed within the content are solely of the author's; and do not reflect the opinions and beliefs of NASSCOM or its affiliates in any manner. NASSCOM does not take any liability w.r.t. content in any manner and will not be liable in any manner whatsoever for any kind of liability arising out of any act, error or omission. The contents of third-party article/blogs published, are provided solely as convenience; and the presence of these articles/blogs should not, under any circumstances, be considered as an endorsement of the contents by NASSCOM in any manner; and if you chose to access these articles/blogs , you do so at your own risk.


Download Attachment

20220325_NASSCOM_Feedback on Framework for Fairness Assessment of AI-ML Systems.pdf

© Copyright nasscom. All Rights Reserved.