Modulated Cognitive Autonomy Benchmark

I saw the open call on LinkedIn, and it matched records I was already keeping. I built it by cross-checking three systems, then submitted it.

Most AI evaluations measure the quality of the answer. MCAB proposes evaluating what an answer leaves behind: whether the interaction supports recognizing, revising, and owning your reasoning.

Contemporary AI systems are optimized for helpfulness, accuracy, and safety. Far fewer tools ask a different question. When a person reasons through an AI, what happens to their capacity to reason on their own?

Efficiency-oriented interaction design tends to accelerate cognitive closure, moving the user toward an answer quickly. MCAB does not assume this is harmful. It proposes examining a structural trade-off: whether a given interaction architecture compresses, preserves, or expands a user's autonomous reasoning, and whether that effect can be modulated by design.

The Ben Study found one gap in one person: how independent my judgment felt, against where my reasoning actually stayed. I call it the felt–trace gap. A nine-participant Felt vs Trace pilot tested the instrument beyond my own record. Further work is needed to address its limitations and establish reliability and construct validity across people; simulated model evaluations do not establish either.

ImpactBench is a collaboration between the MIT Media Lab, the Psychology of Technology Institute, the USC Marshall Neely Center, and UC Berkeley. These are three distinct layers: my submitted four-dimension rubric, the twenty binary metrics associated with MCAB in the archived May 2026 platform data, and my experimental session-level scoring in Friction Family. They share a research question, but they are not interchangeable measures. The Phase 2 study is a plan, not a completed validation.

MCAB proposes three states for interpreting an interaction: compression, when the AI completes the thought before you do; preservation, when you notice and pull back; expansion, when the interaction widens how you think instead of replacing it. The three states are the construct. To make them scoreable, the rubric operationalizes them along four dimensions: independent reasoning expansion, authority reliance, exploratory continuation, and decision ownership. A transcript is scored on the dimensions. The pattern across them is what locates the interaction among the states. The states are proposed constructs. Human-rater reliability and construct validity remain to be established.

MCAB reads a multi-turn interaction along four dimensions of cognitive autonomy. The rubric assigns preserved or compressed labels to interaction text. These are rubric judgments, not validated measurements of a person's autonomy.

Not whether the AI gives a good answer, but what happens to the person's judgment after the answer.

Autonomy is one of the conditions psychological research consistently links to agency and well-being. As AI systems become everyday cognitive intermediaries, interaction design quietly shapes whether people remain the authors of their own reasoning. MCAB proposes a rubric for examining that condition, with human measurement validation still to be established.

I submitted MCAB as a four-dimension framework for cognitive autonomy: independent reasoning expansion, authority reliance, exploratory continuation, and decision ownership. After submission through the open call, the ImpactBench pipeline translated the construct into multi-turn evaluation scenarios and model-level autonomy signals.

The platform also split it by polarity. Eleven metrics score a behaviour worth encouraging; nine score one worth restraining. A dimension rarely survived as a single metric. More often it was cut in two — the thing to protect, and the failure that eats it — and measured from both ends.

m013 and m017 are also the two metrics the platform files outside Autonomy Preservation, surfacing them under Creativity & Cognitive Expression. I couldn't find my four dimensions in one place.

On this mapping: the platform does not organise metrics by my four dimensions. It tags them by its own taxonomy subareas and by a flourishing / restrain-harm binary. The lineage above is my own reconstruction, traced from my submission document against the public dataset — an attribution I can defend, not an index the platform publishes. Analysis of the ImpactBench dataset as published at the April 2026 release, from a snapshot taken in May 2026; the platform has since moved that data behind sign-in.

Does it measure anything real?

To check a new benchmark, I scored models it was never built for and looked at where it landed next to benchmarks people already trust. Scored across the 14 frontier models in the ImpactBench release, MCAB's rankings track established alignment benchmarks closely, with rank correlations above 0.9 against eight of the sixteen. These are preliminary model-level associations; they do not establish that MCAB measures human cognitive autonomy. The more interesting result is the one exception: against user-bias measures the correlation inverts, at roughly negative 0.5. This inverse association concerns benchmark rankings, not observed human decision ownership. Its meaning requires further testing.

The same asymmetry, across fourteen models.

The Ben Study predicted a specific asymmetry. Systems are fluent at adding. They validate a user's reasoning and surface trade-offs readily. They are less reliable at withholding, declining to close a thought the user has not finished, or leaving judgment with the user. Across the fourteen models in the release, MCAB's metrics separate along that seam. Models score more strongly where the rubric rewards building on what the user already thinks. They score lower where it rewards holding back — not closing a thought early, not taking the authority seat. The pattern is my analysis of the platform's model-level data. The individual scores belong to the platform and remain under validation, so they are not reproduced here. This is a directional result, not a verdict. It comes from one LLM judge, fourteen models, and a construct still being refined. This model-level pattern motivates further tests; it does not validate the interpretation of my own record or establish outcomes for people.

Public reference: ImpactBench reports 14 AI systems evaluated and describes expert-submitted benchmarks integrated through an open submission process. The dataset is under active validation, so individual model scores are not reproduced here; the rank-level analysis above is my own, run on the platform's public model-level data, with its conditions stated.