Jennifer Kinne - The Information-Theoretic Imperative

Compression and the Epistemic Foundations of Intelligence

The Lounge always hopes our speakers will make us think differently about what we think we know. And boy, does Jennifer Kinne make us think.

AI systems are routinely described as intelligent: capable of learning, adapting, and generalizing. But intelligence as a design achievement and intelligence as a consequence of physical and informational constraint are very different claims, with very different implications for how we govern these systems and what we expect from them.

This talk presents the Information-Theoretic Imperative (ITI) and its companion principle, the Compression Efficiency Principle (CEP): a two-level framework establishing that any system persisting in uncertain, structured environments must minimize epistemic entropy through predictive compression, and that efficient compression mechanically selects for causal models over superficial statistical patterns through exception-accumulation dynamics. The result is a formal causal chain from survival pressure to reality alignment that reframes intelligence not as an achievement but as a mechanically necessary outcome of persistence under constraint.

For complexity practitioners, the framework offers a substrate-independent account of why systems drift epistemically under distributional shifts (i.e., why a model that worked becomes one that misleads) and what that means for governance. If epistemic alignment is a consequence of the right constraint structure rather than a design specification, then AI governance is not primarily a compliance problem. It is a complexity problem: one that requires attending to the conditions under which systems maintain or lose contact with the causal structure of their environments.

Bio

Jennifer Kinne works at the intersection of information theory, formal epistemology, and AI governance. She has spent over two decades at Harvard University across research operations, regulatory science, and institutional compliance. She is the founder of VeracIQ LLC, CEO of the ITAG Innovation Lab, and a Founding Partner of the Institutional Coherence Initiative. Her paper "The Information-Theoretic Imperative: Compression and the Epistemic Foundations of Intelligence" (arXiv, 2025) proposes that intelligence is not a design achievement but a mechanically necessary outcome of persistence in uncertain, structured environments. It develops the framework presented in this talk.

This event will be recorded. The event link will be published on meetup.com 1 hour prior.

Please note the time slot - this is an afternoon event for those of you on EDT (12:00 pm to 2:00 pm).https://properscience-limited.zoom.us/j/85195718753?pwd=EhZFoM56NlAeDa8BEFWjfHMipwN812.1