Microsoft research cracks the black box of brain-prediction AI with a closed-loop test

For years, the most accurate tools for predicting how the human brain responds to language have been large language models. Feed an LLM the same story a person hears in an fMRI scanner, and the model's internal representations can predict the activity of individual patches of cortex with remarkable fidelity. But this success comes with an unreadable catch: these models are vast collections of learned parameters, not scientific theories anyone can interpret. A model that predicts brain activity tells researchers that a region responds to language, but not what it is actually picking up on, whether it is food, places, numbers, or something else entirely. As black-box models spread, the gap between prediction and understanding has become one of the central problems in computational neuroscience.

In a new paper accepted in Nature Neuroscience, scientists from Microsoft Research, the University of California, Berkeley, the University of California, San Francisco, and Columbia University introduce a framework designed to fix this: generative causal testing (GCT). The method distills brain-prediction models into short, readable accounts of what each patch of cortex responds to. Then it tests those claims by having an LLM write new stories engineered to activate a specific brain area. Subjects hear the stories in the scanner. If the explanation is correct, the targeted region lights up. The result is a way to translate uninterpretable predictive models back into the currency of science: concise hypotheses that can be confirmed or refuted in a follow-up experiment.

How GCT works: explanation, then verification

GCT happens in two steps. First, to generate an explanation, the method starts from a predictive model for a single voxel or region and identifies the short phrases that most strongly drive its predicted response. An LLM then summarizes those words into a concise verbal explanation, often a single phrase such as “food preparation” or “location names.” The second step closes the loop. To build trust in the explanation, GCT uses an LLM to write new stories where each paragraph is carefully constructed to drive a brain region according to its explanation. Three subjects returned to the scanner to read these synthetic stories. If a region's activity to its driving paragraphs was significantly greater than to baseline text, the explanation passed a genuine causal test, not just a correlational one.

Across all three subjects, the core approach held up. The synthetic stories reliably drove their target regions above baseline, confirming that GCT's short explanations capture something the cortex genuinely responds to. The explanations were also most trustworthy where the underlying brain-prediction models were strongest: the more stable the model, the more reliably its explanation could be confirmed in the scanner.

Sharpening the map of place-processing regions

With the method validated on regions whose selectivity was already known, the researchers turned to harder questions. Three neighboring regions involved in processing places have often been treated as functionally similar: the retrosplenial cortex (RSC), the parahippocampal place area (PPA), and the occipital place area (OPA). At first, stories written for one region also activated the others. But by generating differential stimuli, stories designed to switch one region on while keeping its neighbors quiet, the team teased the three apart. For example, RSC responds more strongly to proper noun location names, like Tokyo or Connecticut, rather than general location descriptions. This is the kind of nuanced, region-specific theory that a raw predictive model cannot provide on its own.

Discovery of new prefrontal micro-regions

Beyond known regions, the authors discovered new prefrontal micro-regions that had not been previously mapped. By scanning a grid of candidate locations and keeping only the most stable ones, GCT surfaced these areas tuned to remarkably specific concepts: one selective for dialogue between people (words like “said” or “told”), one for mentions of clock times (“one o'clock”), and one for numeric measurements (“50 feet”). These are distinctions no one had gone looking for; they emerged because the method could propose a hypothesis and immediately test it.

Implications for neuroscience and beyond

GCT's reach goes beyond neuroscience. Researchers in fields from genomics to climate science increasingly face the same dilemma: a model that predicts beautifully but explains nothing. GCT shows that a data-driven model need not be the end of inquiry. It can be distilled into a readable, experimentally testable theory, and that theory can be checked against reality by generating new experiments on demand. For neuroscience specifically, GCT points toward a faster, more hypothesis-rich way of mapping the cortex: an AI system proposes what a brain region might encode, and a closed-loop experiment confirms or rejects it within a single study. The broader lesson is that the rise of black-box models in science does not have to mean the end of human-readable theory. With the right framework, the two can advance together.