about frame
Responsibility belongs to people, not machines.
FRAME (Framework for Responsible AI in Monitoring and Evaluation) gives evaluation teams actionable guidance for using generative AI across the evaluation lifecycle — built around five principles and a checklist that lets teams assess and document responsible AI use, from inception through to how findings are used.
Why FRAME exists
The popular language of "responsible AI" quietly hands moral agency to machines.
FRAME rejects that framing. Drawing on socio-technical systems theory and international governance standards, it locates responsibility firmly with human actors.
AI can reach toward us — but it cannot be responsible.
The thinking behind it
Three ideas shape the framework
The agency attribution problem
Treating AI as a decision-maker creates “moral crumple zones” where humans deflect responsibility onto algorithms. Evaluative judgment — making value judgments about social interventions — cannot be automated.
01
Justice beyond “do no harm”
AI trained on data from unequal societies reproduces those inequalities unless designers actively disrupt them. Responsible use means actively pursuing justice, not just avoiding obvious harm.
02
A socio-technical lens
Technology cannot be understood apart from its social context. A tool that works in one setting can fail in another where culture, language, or relationships differ. FRAME asks evaluators to stay context-aware throughout.
03
Grounded in international governance
Adopting FRAME moves you toward wider compliance
FRAME aligns with and draws on the major AI governance frameworks, so good practice and compliance pull in the same direction.
- NIST AI RMF
- EU AI Act
- OECD
- UNESCO
- ISO/IEC 42001
AI Risk Management Framework
And its Generative AI Profile — Govern, Map, Measure, Manage.
The European Union AI Act
Risk-based classification and human-oversight requirements.
OECD & UNESCO
The OECD AI Principles and UNESCO’s Recommendation on the Ethics of AI.
ISO/IEC 42001:2023
The first certifiable standard for AI management systems.
Questions from AI for MEL course participants shaped the checklist.
Where it came from
FRAME was not written in the abstract
It grew out of real cases — with their messy complexities, unexpected failures, and practical workarounds — and out of questions raised by practitioners who were already using these tools in the field.
- Live cases documented across the Routledge volume From Algorithms to Evidence.
- Questions from participants in The Development CAFE's AI for MEL course.
- Practitioner experience across Asia, Europe, Latin America, and Africa.
The checklist reflects the decisions practitioners actually wrestle with when they deploy AI.
the authors
The people behind FRAME
FRAME is the work of three evaluators who have spent their careers on the practical side of monitoring and evaluation.
Valentine J. Gandhi
The Development CAFE, Indonesia
Founder of DevCAFE and originator of the FRAME approach.
Steffen Bohni Nielsen
National Research Centre for the Working Environment, Denmark
Ready to apply FRAME to your next evaluation?
Download the FRAME checklist, or enrol in the AI for MEL course to put the framework into practice.