Independent reasoning
Can people still solve problems without AI, check false claims and sit with uncertainty?
Research program / interactive simulator
BETA-MIND simulator
Measuring whether repeated AI use improves performance without weakening independent thought, creativity or responsibility.
BETA-MIND in plain English
Can AI make us more capable without making us less able to think for ourselves?
BETA-MIND measures what repeated AI use changes in the person, not only whether the AI completes the task.
It is a doctoral research program and an interactive research model, not a deep-research search engine. The goal is to learn whether people remain able to think, disagree, create and take responsibility after relying on AI over time.
The key difference
Most AI tests ask, “Did the AI give the right answer?” BETA-MIND asks, “After using it repeatedly, what can the human still do alone?”
What changes are we looking for?
Four parts of human agency
The hypothesis is not simply that AI makes people less intelligent. AI may improve immediate results while quietly moving reasoning and decision ownership away from the person.
Handling disagreement
Can people disagree with others, repair conflict and decide when an AI should or should not be trusted?
Creativity
Does AI expand the range of ideas, or make many people produce the same model-typical answers?
Responsibility
Will people challenge unsafe advice and take responsibility for decisions they choose to act on?
What the research will build
01 / SIMULATOR
A virtual society of people and AI assistants
Simulated people interact repeatedly with AI in settings such as a classroom, hiring panel or civic discussion. The simulation speeds up months or years of interaction so researchers can watch for dependency, idea convergence, lost dissent and shifted responsibility.
02 / BENCHMARK
A score for what the human retains
The Human Agency Benchmark reports AI capability and retained human agency separately. A system can produce excellent work and still be a poor result if the person becomes unable to reason or decide without it.
High capability + high agency: the goal
High capability + low agency: dependency risk
How the comparison works
Compare four kinds of AI access over time
Measure before AI use, during repeated use, immediately afterward and after a period without assistance. The question is which abilities last.
- 01
No AI
Establish what people can do independently.
- 02
Always-available AI
Observe what changes when assistance is continuously present.
- 03
Agency-preserving AI
Prompt people to verify, reflect and make the final judgment.
- 04
AI gradually withdrawn
Test what people retain when assistance becomes intermittent or stops.
The ideal outcome: capability gained without human agency lost.
What does BETA-MIND stand for?
Behavioral Ecosystem Testbed for Agents / Multi-Principal Interaction and Network Dynamics.