Measuring what matters: Athena’s framework for responsible AI evaluations

About the Event

AI deployment is rapidly outpacing governance and regulation. Athena Infonomics will showcase its AI Evaluation Framework—a structured approach with four pillars: technical accuracy, human-context validation, attributable developmental impact, and red teaming—for evidence-based, responsible deployment of AI for social good. Drawing on a landscape analysis of AI in agriculture and a live case from Andhra Pradesh, the webinar offers evaluation practitioners, AI developers, and programme managers a replicable framework for responsible AI evaluation.

Speakers

名称	标题	Biography
Dr. Francis Xavier Rathinam	Senior Director - Global MERL, Athena Infonomics	Francis is a seasoned evaluation expert with deep experience in impact evaluations, technical advisory, and M&E systems. He will share lessons from Athena’s AI framework and its integration within evaluation systems, emphasising ethics and practical applications.
Aditi Namdeo	Independent Expert - AI for Social Impact	Aditi will bring a systems-level perspective on responsible AI deployment, drawing on two decades of work across innovation, data governance, financial inclusion, and AI for social impact, including current work on evaluating AI interventions with the Gates Foundation.
Zeba Siddiqui	Senior Consultant - MERL, Athena Infonomics	Zeba is an evaluation professional with demonstrated contributions to several evidence synthesis projects for big data and group-based livelihood interventions in L&MICs. As the co-author of Athena’s AI evaluation framework, she will situate its relevance within Athena’s AI portfolio.

Moderators

名称	标题	Biography
Deepa Karthykeyan	Co-Founder & Partner, Athena Infonomics	Deepa is a leader in multi-disciplinary programme development at Athena, with expertise in systems design and WASH. She will share insights from her experience as the head of Athena’s AI vertical, emphasising gender inclusion and pro-poor approaches in evaluation.

摘要

The webinar highlighted the need to move beyond assessing AI solely on technical performance and toward evaluating real-world effectiveness, impact, and responsible use. Discussions emphasised that AI systems must be assessed within the contexts they operate in, considering user experiences, development outcomes, equity, and potential risks. Speakers noted limitations of existing evaluation approaches, particularly in low- and middle-income countries, and stressed the need for stronger evidence on what works, for whom, and under what conditions. Athena's framework integrates technical, human-centered, and impact-focused dimensions of evaluation, supporting more accountable, inclusive, and contextually relevant AI systems.

Following the webinar, Athena will continue advancing its responsible AI evaluation framework through consultation, testing, and knowledge-sharing. Planned actions include incorporating feedback, refining methodology, and validating the framework across sectors and contexts. Athena also aims to develop practical guidance, tools, and checklists to assess AI systems effectively. A key priority is publishing and disseminating the framework to encourage wider adoption and more consistent evaluation practices. Future efforts will focus on piloting the framework, generating evidence of its effectiveness, and fostering collaboration among researchers, practitioners, governments, and development partners.