Bias testing shall consider the types of bias, types of data and types of models before contextually assessing them for potential bias. The framework provided below guides the process of considering the multiple layers systematically. Components of Bias Testing Framework

Bias is contributed by data, model, outcome, interface, environment, and human actions in which such AI/ML system is used. It could exist in both supervised and unsupervised learning environments in the context of AI/ ML systems. …

The government published a National Strategy for Artificial Intelligence in June 2018 and launched a National AI portal (here) containing a repository of AI initiatives in the country. The national think tank, NITI Aayog, also provided detailed guidance on Responsible AI in 2021 & 2022. In addition, India has established…

Dealing with the risks of behavior biometrics Behavior biometrics pertains to using keystroke cadence, touch dynamics, swipe patterns, mouse movements, clicks, device orientation, and time spent on screen by the user. Combined with other information, including user and device or network identities, these are unique and provide high-level reliability against physiological biometrics (face scanning). Behavior biometrics can…

Framework for evaluating effectiveness of Human-in-the-loop or Human-on-the-loop — HTL in Context Human-in-the loop/ on-the-loop (herein referred as ‘HTL’) is a mechanism through which human intelligence is integrated in using or leveraging machine learning adoption, thereby supporting human discretion with machine activity/ outcomes. HTL is about two key factors (1) Design of Human-Computer Interaction and (2) Autonomy of Decision making. With increased…

GitHub contains pre-trained models that may contain inherent biases, and the absence of model scorecards will contribute to amplifying the bias for users who fork such repos. — Setting the context GitHub is a prominent internet hosting platform for software codes and their version control. GitHub enables its 56 million users (organizations and individuals) to create repositories of their work on the platform for ease of access, version control and for sharing with or without a license (Apache…

Multi-stakeholder feedback has its inherent flaws, hence, its best to treat them as ‘one of the means’ than an ‘as an end’ to establishing and maintaining Ethics in Artificial Intelligence — Setting the context Ethics is a system of accepted beliefs that guide or control behaviour of individuals or groups. Beliefs represent rational, truthful, factual view in some parts and their contrary in some parts in relation to the current time. The accepted beliefs may relate to social philosophies learnt from…

EU AI draft regulations may have an impact on fraud-prediction algorithms and may require review and refinement — Fraud prediction models have existed for a decade or more, in financial services and emerging tech industry including e-commerce, its evolving slowly in multiple other industries for various used cases (eg. warranty fraud prediction) A fraud or cyber security prediction model is a typical prediction model with a specific intent…

Litmus test for travel, mobility, and delivery service platforms — Travel (Ticketing), Mobility (Transport network and ride-hailing platforms) and Delivery service platforms have become a norm in recent years and needless to say that many of us are inherently grateful for their support in the COVID-19 times. With a deep tech foundation, integrated ecosystem view and efforts towards meeting customer…