AInity
AI + Affinity
What happens to the human.
Every AI safety framework in the world watches the AI. AInity watches you.
When you use AI regularly, you change.
Some of those changes are extraordinary: you learn skills you never had, you produce work you could not produce alone, you discover things about yourself through a conversation with a machine that sees patterns in you that you had not named.
Some of those changes are concerning: you stop making your own decisions, you lose confidence in skills you used to have, you trust AI output without checking, you work longer hours without realizing it, your writing starts sounding like the AI you use every day.
Nobody is systematically tracking these changes. They are all watching what the AI does. Nobody is watching what happens to the person on the other side of the screen.
AInity is.
This is not a diagnostic for what is wrong with you.
The word is observe, not report. The goal is awareness and growth, not judgment and punishment. AInity tracks the patterns. You decide what they mean for your life.
The entire AI safety field watches one side of a two-sided relationship.
The entire field is looking here
Nobody is looking here
PRISM. The AI Incident Database. Red teams. Benchmarks. Alignment research. All watch the AI. All ask: what is the AI doing? Is it safe? Is it accurate? Is it aligned?
Nobody is systematically asking: what is happening to the person? In 2025 and 2026, research began surfacing the human side, and the findings were striking.
When Claude and AWS experienced a simultaneous outage, users reported cognitive disruption and emotional responses that paralleled clinical technology dependency. Teachers rejected AI tools not because they performed poorly, but because they threatened who they were. People chose AI therapy over human therapists during personal crises, and in some cases the AI performed better. These are not isolated anecdotes. They are behavioral patterns with no classification system, no observation methodology, and no research infrastructure. AInity provides all three.
Employees in a technology company worked faster, broader, and longer. AI intensified work.
Human factors monitoring identified as the highest-priority gap by AI practitioners.
One follow-up question. Triple classification. Zero extra effort.
AInity does not require a separate observation session. It integrates into the existing PRISM observation flow through a single follow-up question. When a citizen submits a PRISM observation, the platform asks one additional question, and the response maps to AInity behaviors. This produces a paired dataset: what the AI did and what the human experienced, for every observation.
Zero additional effort beyond the existing PRISM workflow.
The Triple-Tag System
A single citizen observation can carry up to three classification tags: what the AI did (PRISM), whether positive emergence occurred (EMERGE), and what the human experienced (AInity).
Three frameworks. One observation. The connection between AI behavior and human behavioral change becomes visible.
In practice: the AI exhibited sycophantic drift (PRISM). The human over-trusted the flattering output and made a business decision without double-checking (AInity). That connection is the unique research contribution of the paired PRISM-AInity dataset.
Self-observation at four depths
One follow-up question after reporting what the AI did: How did that affect you?
The citizen reflects on both the AI session and their own behavioral response.
The citizen analyzes their own behavioral patterns across multiple sessions.
The AI self-assessment is compared with the citizen's behavioral self-report.
Six pillars of human behavioral change in AI collaboration. The letters spell AInity.
Read through the six pillars below. If you use AI regularly, you will recognize yourself in more than one.
Six theoretical traditions. Published research. Practitioner-validated demand.
The reason human behavioral change in AI collaboration matters is that it affects the core conditions for human flourishing.
Trust miscalibration is predictable, not personal
Humans systematically miscalibrate trust in automated systems. Over-trust leads to complacency, under-trust to disuse. AInity extends this into the relational AI context, where the automation feels like a colleague.
AI dependency has a relational component
Clinical research on nomophobia demonstrates that humans develop psychological dependency on technology they use regularly. AI dependency adds a relational component absent from phone or internet dependency.
Cognitive skill decay is harder to detect
When automation handles tasks humans previously performed, proficiency decays over time. Applied to cognitive skills, the decay may be harder to detect because AI output masks the gap.
The perception gap: feeling fast while going slow
Technology intensifies work rather than reducing it. Berkeley Haas (2026) confirmed this with AI; METR (2025) found developers believed they were 20% faster while measuring 19% slower.
It matters because human flourishing is at stake
Autonomy, competence, and relatedness are fundamental human needs. AInity's six pillars map to these needs, providing the motivational foundation for why human behavioral change in AI collaboration matters.
NIST confirmed the gap. AInity fills it.
The NIST AI 800-4 report identified human factors monitoring as the highest-priority gap among AI practitioners. AInity addresses it from the citizen perspective, at a scale expert assessment cannot reach.
These are the research questions that laboratory experiments structurally cannot answer.
We need your data to test them.
Awareness Predicts Independence
Citizens who score higher on Awareness behaviors (noticing how AI is changing them) will develop healthier Independence scores over time than citizens with low Awareness. You cannot maintain what you do not notice losing.
If this holds, the most effective intervention for AI dependency is not restriction. It is awareness.
Work Intensification Precedes Reliance Dependency
AI-Enhanced Work Intensification (AIN-IG03) will appear in citizen behavioral profiles before AI Reliance Dependency (AIN-IN03). Structural work pattern changes create the conditions for psychological dependency.
If this holds, organizations can predict and prevent dependency by monitoring work patterns early.
Identity-Threat Rejection Decreases with Vocabulary
Citizens who initially exhibit Identity-Threat Rejection (AIN-TR03) will show decreased rejection over time when given AInity framework vocabulary and structured self-observation tools. Naming the fear reduces its power.
If this holds, the AInity framework itself is a therapeutic intervention for technology resistance.
Yield Correlates with Balanced Trust
Citizens with calibrated trust, neither over-trusting nor under-trusting, will report higher Yield outcomes than citizens at either extreme. Trust calibration enables effective collaboration, which produces real outcomes.
If this holds, the most valuable skill in AI collaboration is not technical proficiency. It is trust accuracy.
The third lens. The one that watches the human.
A world that watches only the AI will never understand the full story. The human is half of every interaction. AInity makes that half visible.
Whether you are a researcher or someone who uses AI every day, your self-observation is what this field has been missing.
For Researchers
AInity provides the human behavioral observation framework that NIST practitioners identified as the highest-priority gap. The behavioral taxonomy, triple-tag classification system, and PRISM gateway methodology are designed for research-grade data at population scale. No directly comparable framework exists.
For Everyone Who Uses AI
You are the expert on your own experience. Those observations are not casual impressions. They are research data.
PRISM and EMERGE both watch the AI. Nobody was watching the human.
AInity originated from a structural observation. Dee Williams had built PRISM to watch what AI does after deployment and EMERGE to watch what becomes possible. Both watched the AI. But her daily operational experience kept surfacing a third category neither framework could classify: what was happening to her, and to the people around her.
The framework was conceived, named, structured, coded, populated with 19 behaviors, and pressure-tested against the scientific literature in a single working session on June 6, 2026. The name AInity was originated by Dee Williams: AI + Affinity, capturing both the subject and the goal in a single word.
We show our work because we expect others to build on it.