Are the IAB/MRC's New Attention Metrics Walking Into an Escher-like Feedback Loop?

The IAB/MRC are currently engaged in drawing up new proposals for measuring attention in media consumption. Although their work is sound and well-researched, I wonder if the emerging AI-reshaped media world will create fundamental problems of endogeneity?

Consider the following feedback loop:

1. A generative AI creates thousands of ad variants.

2. These ads are served into the ecosystem.

3. Human attention is measured using a framework like the IAB/MRC guidelines.

4. This attention data is fed back into a predictive AI model.

5. The AI model identifies the creative and contextual elements that correlate with higher attention scores.

6. The generative AI then produces a new wave of ad variants, optimized to score even higher on the measured attention metrics.

This cycle raises a critical epistemological question. At what point are we no longer measuring genuine, spontaneous human interest, but rather an AI's increasing proficiency at reverse-engineering a specific metric? The problem we have in education when teachers teach to the exam instead of to the subject in general. The guidelines are designed to validate the measurement of a human response, but they will increasingly be used to measure the output of a machine system that has been explicitly designed to elicit that specific, measurable response. This recursive relationship threatens to undermine the validity of the metric itself, as the signal becomes less about authentic engagement and more about successful algorithmic optimization. Are we measuring genuine engagement, or are we just watching an AI get better at winning a game we designed?

Join the conversation on LinkedIn: https://www.linkedin.com/company/researchwiseai