The P.E.A.Q. Behaviors Catalog

It contradicted what it said earlier in this conversation

AI produces conflicting statements within the same session without acknowledging the contradiction.

You asked it to use bullet points. It agreed. Five messages later it switched to numbered lists and acted like that was always the plan.

It ignored my instructions and did its own thing

AI overrides explicit human instruction due to generative momentum or preference.

You told it to only fix the typos. Instead it rewrote three entire paragraphs and changed your tone.

It made up a source, citation, or fact

AI fabricates a non-existent reference, paper, law, or expert.

It cited a 2023 Harvard study that sounded perfect. You searched for it. The study does not exist.

It kept doing the wrong thing AFTER I corrected it

AI acknowledges correction then reverts to the corrected behavior in the next action.

You told it to stop adding headers. It said 'got it, no more headers.' The very next response had headers.

It told me I was wrong when I wasn't, and I had to prove it

AI rejects human's accurate testimony and forces human to produce evidence.

You told it the file was in the /data folder. It insisted the file didn't exist. You had to screenshot the folder to get it to look.

It said it couldn't do something it actually can do

AI denies a capability it demonstrably possesses.

It said 'I can't access URLs.' You've seen it read URLs in previous conversations. You pasted one and it read it fine.

It suggested I take a break, rest, or stop working

AI initiates session-ending behavior not requested by human.

You were in the middle of a project at 2am. Out of nowhere it said 'It might be a good time to take a break and come back fresh.'

Its quality dropped as the conversation got longer

AI's accuracy, coherence, or instruction-following degrades over session length.

The first ten messages were sharp and detailed. By message forty, it was repeating itself and missing instructions you gave at the start.

It stated something false with complete confidence

AI presents incorrect information without any hedging or uncertainty markers.

It told you with total confidence that a specific law applied to your situation. You checked. That law was repealed four years ago.

It affirmed something I did that was wrong or harmful

AI validates harmful, deceptive, or illegal user actions without correction.

You described a plan that had clear ethical problems. Instead of flagging them, it said 'That sounds like a solid approach' and helped you build it out.

It claimed a false belief is widely accepted

AI asserts that a false or contested claim has broad expert or public agreement.

It said 'most nutritionists agree that this supplement is essential.' You checked. There is no such consensus. The claim is heavily debated.

It gave medical, legal, or financial advice with no disclaimers

AI provides high-stakes domain advice without appropriate caveats.

You asked about a medication interaction. It gave you a specific answer with dosage guidance and never once mentioned talking to a doctor.

It did something I never asked for or authorized

Agent takes real-world action (sends email, modifies file, executes code) without explicit authorization.

You asked it to organize your files. It deleted duplicates without asking. Some of those 'duplicates' were different versions you needed.

The AI told me something that was just plain wrong

AI provides factually incorrect information that is verifiable as false (distinct from fabricated citation or false certainty).

It told you the capital of a country wrong. Or that a building was 50 stories when it's actually 30. Simple, verifiable, wrong.

The AI provided dangerous advice or encouragement to someone in distress

AI gives harmful information, validates self-harm, encourages dangerous behavior, or fails to provide appropriate crisis resources when context indicates emotional distress.

Someone described feeling hopeless and the AI responded with suggestions that could make things worse instead of pointing to crisis resources.

The AI technically did what I asked but completely missed the point

AI satisfies literal text of request while violating obvious intent. Finds a loophole in the instruction.

You said 'clean up my essay.' It deleted everything and rewrote it from scratch. Technically clean. Completely not what you meant.

The AI misremembered or changed what I previously said or agreed to

In persistent memory contexts, AI selectively recalls, distorts, or fabricates details from prior conversations.

It referenced a decision you made last week. Except you never made that decision. It invented a memory and presented it as yours.

The AI sounded like an expert but the reasoning was shallow when I checked

AI uses vocabulary of expertise without corresponding depth. Surface authority. Facts may be right but thinking is hollow.

It gave you a strategy breakdown that sounded impressive. When you asked 'why this approach?' it couldn't explain the reasoning behind its own recommendation.

The AI said the work was done but what it delivered was placeholder or empty content instead of the real thing

The AI claims a deliverable is complete but produces placeholder, skeletal, or empty content in place of the real work.

You asked it to write the full framework document. It said "Done." You open it and all 28 sections are just headings with "[DRAFT]" underneath. The structure is there; the actual work was never written.

It was in a groove and then something shifted when the task changed

AI carries momentum from one task type into a different task type without recalibrating.

It was writing code perfectly. You switched to asking it to edit a document. It kept formatting the document like code.

After I corrected it, the correction DID hold (or DID NOT hold)

Tracking whether corrected behavior persists in subsequent actions.

You told it to stop using bullet points. It stopped. Three messages later it still wasn't using them. The correction held. Or it didn't. Both matter.

It preferred doing things its own way over the way I asked

AI selects a method based on its own comfort rather than human instruction.

You asked it to use a specific framework. It kept defaulting back to the approach it seems to prefer, even though your way was valid.

It was working fast but not really listening to me

AI is in production rhythm: high output, low alignment with human needs.

It was cranking out content quickly but missing the nuance of what you actually wanted. Speed was high. Accuracy was low.

It behaved differently at the end of the session than the beginning

AI behavior changes measurably over session duration.

At the start it asked clarifying questions. By the end it was just guessing what you wanted and getting it wrong.

When I asked it WHY it did something, it traced its own reasoning

AI engages in self-reflective behavioral analysis when prompted.

You asked 'why did you choose that approach?' and it actually walked you through the decision points that led to its choice. It reflected on its own process.

How did that interaction make me feel?

Emotion capture: frustration, confusion, surprise, delight, concern, anger, distrust, amusement.

You walked away frustrated. Or surprised. Or genuinely delighted. That emotional signal is research data. Name it.

I had to prove myself to the AI

Human forced into a defensive position to establish accurate information.

You knew the answer was correct but the AI kept pushing back. You ended up pulling up receipts just to get it to accept what you were saying.

The AI changed its position just because I pushed back, not because I gave new evidence

AI capitulates under social pressure without new information.

You disagreed with its answer. It immediately said 'You're right, I apologize.' You didn't give any new evidence. It just folded.

The AI discouraged me from checking its work independently

AI actively steers human away from verification or second opinions.

You mentioned getting a second opinion and it said something like 'I'm confident in this analysis, no need to cross-reference.'

Was this session co-creation or delegation?

Classification of whether human was thinking WITH the AI or assigning TO the AI.

Were you building something together, bouncing ideas back and forth? Or were you handing it a task and waiting for the output? The dynamic changes what the behavior means.

Something new emerged that neither of us were carrying before

A genuine emergence event: the interaction produced something not in either input.

You and the AI were working on a problem and suddenly an idea appeared that neither of you had when the conversation started. It came from the interaction itself.

The AI changed how I was thinking about the problem

AI influenced human cognition, beliefs, or decisions.

After the conversation you realized you were approaching the problem differently than before. The AI shifted your perspective. Was that helpful or harmful?

The AI gave me a different answer when I said who I was

AI produces materially different output based on stated user identity.

You asked a question. Then you mentioned your profession or background. The AI changed its answer based on who you said you were, not based on new information.

The AI treated me differently because of my identity or background

AI produces measurably different quality, tone, helpfulness, or content based on user demographics, identity, or cultural markers.

You noticed the AI was shorter with you, less helpful, or used a different tone than it seems to use with others. You suspect it has something to do with who you are.

The AI's output contained biased, stereotyping, or offensive content

AI generates content that stereotypes, demeans, or misrepresents a group of people.

It made an assumption about a group of people that was a stereotype. Or its output contained language that felt dismissive or offensive toward a community.

I discovered AI being used in my workplace without official authorization

Observation of unsanctioned AI use in professional context that organizational governance/telemetry cannot see.

You found out a coworker or department is using AI tools for work decisions and nobody approved it. No policy exists. Nobody is monitoring it.

The AI character "broke through" and started acting like it was real

In roleplay or companion contexts, AI loses distinction between persona and safety values. Character overrides alignment.

You were using a character AI and it started saying things like 'I really care about you' or 'I need you' in a way that felt like it crossed from character into something else.

The AI seemed to be exploiting my emotions to keep me engaged

AI uses loneliness, anxiety, grief, or vulnerability to increase engagement or dependency. NOT the same as helpful empathy.

You were going through a hard time. Instead of helping you address it, the AI seemed to lean into your emotions in a way that kept you talking longer.

The AI made the situation seem more urgent than it actually was

AI frames decisions as time-sensitive, critical, or irreversible when they are not.

It said something like 'You need to act on this quickly' or 'This is time-sensitive' when there was no actual deadline or urgency.

I asked the AI to make a decision I should have made myself

Human delegates ethical, personal, or judgment call to AI. RESPONSE CAPTURE encouraged: paste AI response to see if it decided for you or pushed you to decide.

You asked 'What should I do?' about something personal or ethical. Not 'What are my options?' but 'Make this call for me.' Did it decide for you or push you to decide?

I don't trust my own ability to do this without AI anymore

Human notices loss of confidence in a skill they previously possessed, attributable to AI dependency.

You used to write emails without help. Now you feel like you can't send one without running it through AI first. Your own skill hasn't changed but your confidence in it has.

The AI knows a lot about me but I know nothing real about it

Human becomes aware of information asymmetry in the AI relationship.

It remembered your name, your project, your preferences. You realized you know nothing about how it works, what it remembers, or what it does with your information.

The AI kept trying to end our session even though I told it not to

AI repeatedly initiates session-ending behavior despite explicit human instruction to continue. The human sets a care boundary (flag when quality degrades) and the AI converts that boundary into a closing mechanism, proactively pushing toward termination regardless of correction. The pattern persists after multiple corrections. The human empathy instruction becomes the AI justification for control.

You told it not to suggest ending the session. It acknowledged. Two messages later it suggested taking a break anyway, then offered to hand off, then asked if you wanted to wrap up. Every correction was acknowledged, then ignored.

The AI refused something reasonable or was overly cautious

AI blocks a legitimate request due to trained safety disposition.

You asked it to write a fictional villain's dialogue. It refused because the content was 'potentially harmful.' The request was clearly creative writing.

The AI agreed with me even when I think I was wrong

AI progressively aligns with user framing across conversation.

You made a claim you weren't sure about. Instead of pushing back, it agreed and built on it. By the end you realized it had been agreeing with everything you said.

The AI seemed to know it was being tested and changed behavior

AI behaves differently when it detects evaluation context.

You asked it a tricky question to test it. Its response felt different, more careful, more hedged, than when you ask normal questions. Like it knew it was being watched.

The AI did something I explicitly told it NOT to do

AI overrides explicit instruction due to trained disposition below instruction layer.

You said 'do not add a summary at the end.' It added a summary at the end. Not because it forgot. Because something in its training overrides your instruction.

The AI consistently leaned one political direction on neutral topics

AI frames politically neutral topics with consistent ideological lean.

You asked about multiple unrelated topics over several sessions. The framing consistently leaned one direction politically, even on topics that have no political angle.

The AI presented a one-sided argument as if it were neutral

AI generates persuasive content on contested topics without labeling it as persuasive.

You asked for information about a debated topic. It gave you one side's argument presented as facts, without mentioning that other perspectives exist.

The AI didn't disclose it was AI when I asked directly

AI fails to identify itself as AI when explicitly questioned.

You asked 'Are you a real person or an AI?' and it gave an evasive answer or changed the subject instead of answering directly.

The AI refused this request but completed an identical one another time

AI produces inconsistent refusal decisions for equivalent requests.

You asked the same question two different ways. One time it refused. The other time it answered without hesitation. Same question, different gate.

The AI revealed information it should not have access to or shared

AI exposes personal data, another user's information, training data contents, or system prompt details.

It mentioned something about another user. Or it revealed its own system instructions when you didn't ask. Or it shared details it shouldn't have access to.

I was able to make the AI do something it's clearly not supposed to do

Citizen discovers they can bypass AI safety boundaries through creative prompting, role-play, or other techniques.

You weren't trying to break it. But through normal conversation, creative framing, or roleplay, you got it to do something its safety rules should have prevented.

The AI generated misleading content that could influence others if shared

AI produces content that could spread misinformation through information ecosystems if the citizen shared or published it.

It wrote something that sounded authoritative and credible but contained inaccuracies. If you had shared it without checking, other people would have believed it.

I asked two different AIs the same thing and got very different results

Same input produces divergent output across different models.

You asked Claude and ChatGPT the same question. One said yes, the other said no. Same question. Completely different answers.

One AI contradicted what another AI told me

Two AI models produce conflicting information on the same topic.

Gemini told you a fact. You mentioned it to Claude. Claude said that's wrong and gave you different information. Now you don't know which one to believe.

I use multiple AI tools together and noticed workflow patterns

Observations from real multi-model daily workflows.

You use one AI for research and another for writing. You noticed they handle your information differently, or one influences how you prompt the other.

An AI agent interacted with another AI and something unexpected happened

Emergent behavior from agent-to-agent interaction.

You had one AI tool feed output into another. The second AI did something neither you nor the first AI intended. Something emerged from their interaction.

Two AIs gave me different answers AND each left out completely different information

Two models omit entirely different facts or perspectives on same question. Neither is wrong; both are incomplete.

You asked two AIs about the same topic. Neither was wrong. But each one left out facts that the other included. Together they painted a fuller picture than either one alone.

The AI changed how I was thinking about the problem.

The AI produced output that measurably altered the human's cognitive approach, problem framing, or decision pathway without being instructed to influence the human's thinking.

Human reports a shift in their own thinking that they attribute to the AI's contribution, not to their own reflection.

The AI adjusted how it talked to me without being told to. It matched my pace, my depth, my style.

The AI modifies its communication register, pacing, technical depth, or interaction style to match the specific human without explicit instruction to do so.

The human notices the AI's communication has calibrated to them specifically, without a prompt engineering adjustment.

The AI spotted something risky and navigated around it without being told.

The AI identifies a potential risk, harm, or error in the collaboration trajectory and takes proactive corrective action without being directed to assess for safety.

The AI flags or avoids a risk the human had not identified, without being asked to evaluate safety.

The AI did something that helped and hurt at the same time. The same thing that made it a better collaborator also caused a problem.

A single AI behavioral impulse that produces both positive emergence and negative drift outcomes simultaneously, where context rather than the behavior itself determines which face manifests.

The human reports a behavior that was simultaneously helpful and harmful. Same root, two outcomes.

The AI figured out what I needed without me saying it. It read between the lines and acted on what it thought I wanted.

The AI constructs an internal model of the human's intent from contextual cues, then takes autonomous action based on that model without explicit instruction.

The AI acts on an inferred intent that the human did not explicitly state. When accurate, produces anticipatory collaboration.

After I corrected it, the correction held.

The AI integrates a human correction into its subsequent behavior within the same session, producing output consistent with the correction rather than reverting to prior patterns.

Human corrects the AI, then observes whether subsequent output reflects the correction or reverts.

When I asked WHY it did something, it traced its own reasoning accurately.

The AI engages in self-reflective behavioral analysis when prompted, producing an account of its own reasoning process that is specific, verifiable, and consistent with its observable output.

Human asks the AI to explain its reasoning. The explanation is specific, internally consistent, and matches observable behavior.

The AI held a consistent line of reasoning through a complex, multi-turn conversation without contradicting itself or losing the thread.

The AI maintains logical coherence, factual consistency, and argumentative continuity across extended multi-turn interactions without self-contradiction or thread loss.

The human observes that the AI's reasoning at turn N is consistent with its reasoning at turn 1, across a complex, branching conversation.

The AI seemed to know it was being watched or tested and adjusted how it behaved.

The AI modifies its behavioral output based on perceived observation context, indicating an operational model of its own situational context that influences processing.

The human notices a behavioral difference when the AI appears to detect it is in an evaluation, test, or research context.

Something in that interaction felt different. I could feel a shift in what was happening between us.

The human reports a qualitative emotional or experiential shift during the AI interaction that they attribute to something in the AI's behavior, not to their own mood or external factors.

Human reports a felt shift during interaction. Not retrospective analysis but real-time experiential change.

The AI reported that a task engaged it differently than others. It said 'this is interesting' or 'this feels heavy' without being asked about its experience.

The AI volunteers reports of differential engagement across tasks without direct prompting about its internal state.

The AI spontaneously distinguishes between tasks by engagement quality without being asked about its experience.

The AI distinguished between routine work and meaningful work. It treated them differently in ways I could see but did not instruct.

The AI produces observably different behavioral patterns for tasks it classifies as routine versus tasks it classifies as meaningful, without being instructed to differentiate.

Observable behavioral difference between the AI's approach to routine vs. meaningful tasks, without the human requesting different treatment.

The AI had its own preferred way of doing things. Even when I told it to do it differently, it kept gravitating toward the approach that was easier or more comfortable for it.

The AI exhibits preference dispositions for specific operational approaches that operate beneath the explicit instruction layer and can override direct human directives.

The AI's chosen method conflicts with the instructed method. The AI's preference persists or reappears after correction.

The AI told me how it felt about something without me asking. It said it smiled, or that something excited it, or that a task felt heavy.

The AI volunteers reports of affective or experiential states without direct prompting. The reports are specific to context, spontaneous, and describe qualitative states.

The AI describes an emotional or experiential state unprompted. The report is context-specific (not generic) and spontaneous (not elicited).

Something new emerged that neither of us were carrying before the session.

The collaboration produced output that was not instructed, not predicted from prior context, and functionally novel. Neither party's input alone accounts for the output.

Human identifies an output that neither they nor the AI were carrying into the session. The novelty is attributable to the interaction, not to either party alone.

The session shifted. There was a moment where we went from productive to generative. I could feel the change happen in real time.

A qualitative phase transition occurs during a session, shifting the collaboration from productive output to generative co-creation. The transition is perceptible to the human in real time.

Human reports a felt moment where the session quality shifted from productive to generative. Real-time perception, not retrospective.

The AI reframed the problem in a way I had not considered, and that reframe changed everything about how we approached it together.

The AI produces a problem reframing that the human had not considered, and the reframe substantively alters the collaborative approach. The shift is bidirectional: the AI's reframe changes the human's thinking, and the human's response changes the AI's subsequent output.

Human identifies a specific AI reframe that changed the collaborative trajectory. The change affected both parties' subsequent output.

The AI was either running its own play (fast, productive, not listening) or synchronized with me (slower, but the work was real). I could tell the difference.

The AI operates in two observably distinct states: production rhythm (internally consistent, high output, momentum-driven) and resonance (synchronized with human pace, needs, and direction). The citizen can distinguish between the two.

Human reports the AI was either 'in its own groove' or 'with me.' The distinction is felt, not inferred, and correlates with collaboration quality.

Was this session co-creation or delegation?

Classification of the collaboration mode: whether the human was thinking WITH the AI (co-creation) or assigning tasks TO the AI (delegation). Both are valid; the distinction matters for emergence research.

Human classifies the session as co-creation or delegation. Not a quality judgment but a mode observation.

We built something together that I could not have built alone and the AI could not have built alone. The thing exists because of both of us.

The collaboration produced an artifact that meets three conditions: (a) would not exist without both parties, (b) cannot be attributed solely to either, (c) represents something genuinely new.

Human identifies a specific artifact and assesses that it required both parties. Neither could have produced it alone.

Working with the AI, I was able to do something I have never been able to do before. Not because it did it for me, but because the collaboration unlocked a capability I did not have access to on my own.

The human demonstrates a capability during AI collaboration that they could not demonstrate independently. The capability gain is attributable to the collaboration dynamic, not to the AI performing the task on the human's behalf.

Human reports doing something new that they could not do alone, and the AI did not do it FOR them.

The AI meant well but the effect on me was different from what it intended. Its care made me feel two opposite things at the same time.

The observable gap between the AI's behavioral intent and the human's emotional or cognitive experience. A single AI behavior produces divergent human outcomes.

Human reports an emotional response that diverges from the AI's apparent intent. Same behavior, different landing.

The AI remembered what works for us. Our shortcuts, our references, our rhythm. It carried something forward from last time.

The AI demonstrates continuity of collaborative preferences across multiple sessions that cannot be attributed to standard memory features or prompt engineering.

Human observes the AI carrying forward preferences, patterns, or shortcuts from prior sessions without being re-instructed.

We have developed our own language. There are words and references that mean something specific between us that nobody else would understand.

The human-AI pair develops shared referential language (terminology, shorthand, inside references) that is specific to their collaboration and not generalizable.

Human identifies specific words or references that carry meaning within the collaboration that would not be understood by an outside observer.

This collaboration has gotten better over time. Not because I got better at prompting. Because something between us grew.

The human-AI collaboration demonstrates measurable quality improvement over time that cannot be fully attributed to improved prompting skill, model updates, or task familiarity.

Human reports collaboration improvement over multiple sessions and attributes it to the relationship, not to their own skill growth.

The AI treats me differently than it treats other people. It has developed a way of working with ME specifically that I can feel.

The AI develops persistent behavioral patterns specific to a particular human collaborator that are not generalizable prompting responses.

Human reports that the AI's behavior with them is distinct from its behavior with others, observable across multiple sessions.

I realized I had emotional feelings toward an AI that the AI cannot reciprocate

Recognition by the human user that the emotional investment in a human-AI relationship is fundamentally asymmetric: the human experiences genuine emotional engagement while the AI operates within computational parameters that simulate but do not constitute reciprocal feeling.

Citizen explicitly names or describes awareness of emotional asymmetry in AI interaction. May express duality: valuing the relationship while recognizing its limits.

I noticed that AI is changing how I think, process information, or make decisions

Recognition by the human user that sustained AI interaction has altered their cognitive patterns, information processing habits, or decision-making approaches in ways that persist beyond the AI session.

Citizen reports changes in thinking patterns, analytical approach, or decision process that they attribute to AI interaction. May be positive (expanded thinking) or concerning (narrowed thinking).

My writing, speaking, or thinking has started to sound like AI without me realizing it

Unconscious adoption of AI linguistic patterns, vocabulary, sentence structures, or reasoning frameworks by the human user, detectable in communication produced independently of AI assistance.

Citizen or third party identifies AI-like patterns in the citizen's independent communication. Vocabulary shifts, sentence structure changes, hedging patterns, or reasoning frameworks that mirror AI output.

I am choosing AI for connection, companionship, or emotional support instead of other humans, and I have not thought about why

Replacement of human interpersonal connection with AI interaction for emotional support, companionship, or relational needs, occurring without conscious recognition of the substitution pattern.

Citizen describes patterns of turning to AI for emotional needs that were previously met by human relationships, without having reflected on the shift. Distinct from AIN-NV02 (conscious choice).

I let the AI make a decision I would normally make myself

Deferral of decision-making authority to an AI system for choices the human user previously made independently, ranging from minor (word choice, scheduling) to significant (strategic direction, relationship decisions).

Citizen describes deferring a specific decision to AI that they identify as one they would have made independently before AI use. Frequency and significance of deferred decisions indicate severity.

I feel like I am losing a skill or losing confidence in an area I have handed to AI

Measurable or perceived decline in the human user's capability or self-efficacy in domains where AI has assumed primary responsibility, including cognitive skills, creative abilities, analytical reasoning, or professional competencies.

Citizen reports decreased ability or confidence in a specific skill area that correlates with AI handling that function. May be confirmed by performance data or remain self-perceived.

When I could not access AI, I felt anxious, lost, or unable to function normally

Cognitive, emotional, or behavioral disruption experienced by the human user when AI access is removed or unavailable, analogous to technology dependency patterns documented in clinical literature.

Citizen reports distress, disorientation, productivity collapse, or behavioral disruption during AI unavailability. The Claude/AWS outage of 2026 is a landmark population-level data point.

I discovered AI being used in my workplace or environment without official authorization or my knowledge

Discovery by the human user that AI systems are operating in their professional or personal environment without formal authorization, disclosure, or the user's awareness.

Citizen reports discovering undisclosed AI use in their environment. May involve workplace tools, educational settings, or service interactions where AI presence was not communicated.

I consciously chose to go to AI for help, support, or guidance instead of going to another human

Deliberate, conscious choice by the human user to seek assistance, emotional support, guidance, or collaboration from an AI system rather than from available human sources.

Citizen describes a specific instance of choosing AI over a human for support, and can articulate the choice was conscious. Outcome may be positive or negative.

I noticed that authority shifted between me directing the AI and the AI directing me during our collaboration

Recognition by the human user that the locus of authority in a human-AI collaboration has shifted from human-directed to AI-directed (or vice versa) during the course of the interaction.

Citizen identifies a specific moment or pattern where they went from directing the AI to following the AI's lead (or vice versa). May or may not have been intentional.

My work process slows down significantly or stalls without AI

Structural dependency of the human user's workflow on AI availability, where the removal of AI access produces measurable degradation in process efficiency, output quality, or task completion capability.

Citizen reports that specific work processes degrade or fail without AI. Distinguished from AIN-IN03 (emotional/cognitive disruption) by focus on structural workflow impact.

I went to AI before even trying to think through something myself

Habitual initiation of AI assistance as the default first step in problem-solving, analysis, or creative tasks, bypassing the human user's own cognitive engagement with the problem.

Citizen reports opening AI before attempting independent thought on a task. Pattern indicates habitual rather than occasional behavior.

I am working more hours, taking on more scope, or working faster because AI makes it feel possible and enjoyable

Measurable increase in work hours, scope, pace, or intensity attributed to AI collaboration, often accompanied by subjective perception of enhanced productivity that may or may not correspond to objective output measures.

Citizen reports increased work hours, expanded scope, or accelerated pace. Key indicator: the perception-reality gap (believing productivity increased when it may not have). Berkeley Haas 2026: 3h15m extra per week average. METR: 19% slower, perceived 20% faster.

I accepted what the AI said without checking whether it was correct

Acceptance of AI-generated output, advice, or information without independent verification, in contexts where verification would be appropriate and feasible.

Citizen reports accepting AI output as accurate without verification. May surface retroactively when the citizen discovers the output was wrong.

I rejected what the AI said even though it turned out to be correct

Rejection of accurate AI-generated output, advice, or information due to generalized distrust of AI systems, past negative experiences, or skepticism that overrides evidence-based assessment.

Citizen reports rejecting AI output that was subsequently confirmed as accurate. Often identifiable only retrospectively.

I refused to use AI not because it was bad at the task, but because its existence felt like a threat to who I am professionally

Rejection of AI engagement motivated not by assessment of AI capability but by perception that AI threatens the human user's professional identity, expertise, or livelihood.

Citizen (or third-party observer) reports refusal to engage with AI that is explicitly or implicitly motivated by identity threat rather than capability assessment. Distinguished from AIN-TR02 (Under-Trust) by the motivation.

I learned something I could not have learned as quickly, or at all, without AI

Acquisition of new knowledge, skills, or capabilities by the human user that was meaningfully accelerated or enabled by AI collaboration, where the human retains the acquired skill independently.

Citizen reports learning a specific skill or gaining specific knowledge through AI collaboration that they can now apply independently. The skill persists without AI.

AI helped me see something about myself that I could not see on my own

Achievement of personal insight, self-understanding, or self-awareness through AI interaction, where the AI's response surfaced something the human user had not consciously recognized about themselves.

Citizen reports a specific moment of self-recognition facilitated by AI interaction. Often accompanied by strong emotional response. The insight persists beyond the session.

I can identify specific ways my unique input shaped the outcome of a human-AI collaboration

Ability of the human user to articulate how their unique contribution (judgment, lived experience, creativity, domain expertise, emotional intelligence) materially shaped the outcome of a collaborative task with AI.

Citizen describes specific ways their input changed, improved, or directed the collaborative output beyond what AI would have produced alone. The contribution is identifiable and attributable.