.png){kind=spacer}
CLEAR Dialogue: AI and Human Judgment in Monitoring and Evaluation
- CLEAR-AA
Exploring the opportunities, limitations and implications of artificial intelligence for evaluation practice
On the 14th August 2026, CLEAR-AA hosted the inaugural CLEAR Dialogues, bringing together evaluation practitioners and stakeholders for an engaging conversation on one of the issues increasingly shaping the M&E field: the role of artificial intelligence (AI) in monitoring and evaluation (M&E), and the continued importance of human judgement.
The dialogue, themed “AI and Human Judgment in Monitoring and Evaluation,” featured Dr Steven Masvaure, Senior Researcher and M&E Technical Specialist at CLEAR-AA, and Dr Mzamani Mdaka, who leads the Integrated District Improvement Programme at the National Education Collaboration Trust. The discussion was facilitated by CLEAR-AA and ClearPath MEL, an emerging organization focused on strengthening evidence-driven decision-making and improving development outcomes through monitoring, evaluation and learning. This created space for both seasoned and emerging voices to interrogate what AI means for the future of evaluation.
AI as a tool for better evaluation
Dr Masvaure opened the discussion by challenging the idea that AI should be viewed primarily as a threat to evaluators. Instead, he positioned AI as another tool in the evaluator’s toolbox, with the potential to support tasks such as data processing, analysis and synthesis, freeing evaluators to focus more deeply on interpretation, sense-making and judgement.
However, the value of AI depends on how it is used. While AI can process large volumes of information and identify patterns, it cannot independently determine what evidence means within a particular social, cultural or programme context. For Dr Masvaure, this distinction is critical: AI can support evaluation, but it cannot replace the evaluator’s responsibility to make judgements about value, relevance and meaning.
This point resonated strongly with participants. One participant reflected that “evaluation is about valuing and making judgements” and that technological tools should not replace the human role in valuing, but rather help evaluators become more efficient and effective. Another participant reinforced the point, noting that human value judgement remains critical.
The discussion also considered how AI could contribute to more adaptive approaches to M&E. Rather than waiting until the end of a programme to identify what worked or did not work, AI-supported tools could potentially help organisations identify emerging patterns and issues during implementation, enabling more timely learning and decision-making.
Are we ready for AI?
Dr Mdaka shifted the conversation towards an equally important question: having access to AI does not necessarily mean that organisations are ready to use it.
He highlighted the broader conditions required for responsible AI adoption, including data quality, organisational capacity, appropriate infrastructure, governance, ethics and accountability. AI should not be introduced simply because the technology is available; rather, organisations should first understand the problem they are trying to solve and then determine whether AI is an appropriate response.
This was particularly relevant to the South African and broader African context. Participants raised concerns around data quality, digital inequality, regulatory obligations and the rapidly changing risk environment, asking how AI can enhance decision-making while ensuring that human judgement continues to provide the contextual understanding and ethical oversight necessary for fair and trustworthy outcomes.
The conversation also highlighted the changing skills required of evaluators. Participants noted the growing importance of data science literacy, working with large and unstructured datasets, critically assessing machine-learning outputs and developing prompt-engineering skills. As one participant put it, future practitioners will need to be able to use AI productively while also quality-assuring its outputs
AI, bias and the African evaluation context
One of the strongest threads running through the dialogue was the question of context and whose knowledge is represented in AI systems.
Dr Masvaure connected this to ongoing conversations around Made in Africa evaluation, decolonisation and Indigenous Knowledge Systems. If AI tools are trained predominantly on knowledge and data that do not adequately represent African realities, there is a risk that they may reproduce existing biases rather than help address them.
Participants pushed this issue further, asking how evaluators can detect and correct algorithmic bias where African and marginalised populations are under-represented in training data. Others questioned how much progress Africa is making in decolonising evaluation when many education systems and curricula remain Western-centric.
These questions point to an important challenge for the evaluation community: AI adoption cannot simply be about becoming technologically advanced. It must also be about ensuring that the technologies being used are fit for purpose, contextually relevant and capable of reflecting the diversity of the communities and realities that evaluation seeks to understand.
Ethics, transparency and the future of evaluation
The discussion further explored the ethical implications of AI use. Participants raised concerns around AI ethical washing, asking how evaluators can ensure that claims of responsible AI use translate into meaningful practice.
Transparency was another key concern. One participant asked how transparent AI-assisted analysis should be in evaluation reporting, including whether evaluators should disclose the extent to which AI was used in their analysis and reporting. Questions of work ownership and attribution were also raised as AI becomes increasingly embedded in professional practice.
These exchanges reinforced the need for evaluators and commissioners to think proactively about responsible AI use - including expectations around transparency, data protection, accountability and the disclosure of AI-assisted work.
A People-Centred Future
The dialogue ultimately pointed towards a future in which AI and human judgement work alongside one another rather than in competition. AI may increase efficiency, support analysis and help evaluators work with increasingly complex information, but people remain responsible for interpreting evidence, understanding context, exercising ethical judgement and making decisions.
The inputs of participants demonstrated that these are not abstract questions for the future, but they are already emerging in the everyday work of evaluators. The discussion therefore urged evaluators to engage with these shifts now and play an active role in shaping their influence on practice.