Contributors
0 of 1,000,000
Observations
0 of 1,000,000,000
Start Observing →
PILLAR A: AWARENESS
AIN-AWFoundation Pillar

Pillar A: Awareness

The study of whether humans recognize how AI is changing their cognitive patterns, emotional responses, and behavioral habits.

Every AI safety framework in the world watches the AI. PRISM watches what the AI does wrong. EMERGE watches what the AI makes possible. Red teams test what the AI can be tricked into doing. Benchmarks measure what the AI is capable of. Every lens, every framework, every measurement system points at the same target: the machine.

AInity points the mirror the other way. It watches the human.

Awareness is the foundation pillar of AInity. It asks the question that most AI users have never asked themselves: is AI changing you, and do you notice it happening? Not whether AI is useful. Not whether AI is dangerous. Whether AI is reshaping how you think, how you feel, how you communicate, and who you turn to, and whether you are aware of the reshaping while it occurs.

Without awareness, every other AInity pillar is compromised. The human who does not notice that their writing has started sounding like ChatGPT cannot make an intentional choice about it. The human who does not recognize that they have developed emotional feelings toward an AI cannot calibrate those feelings against reality. The human who is choosing AI for connection instead of other humans without realizing it cannot ask themselves whether that substitution is serving them.

Awareness is not judgment. The AInity framework does not diagnose what is wrong with humans who use AI. It observes how humans change when they work alongside AI. Some of those changes are positive. Some are concerning. All of them are worth noticing. The Awareness pillar tracks whether the noticing itself is happening.

The National Institute of Standards and Technology (NIST AI 800-4, 2026) identified human factors monitoring as the highest-priority gap among AI practitioners. Existing monitoring tools focus on model performance, security, and compliance. Practitioners reported needing tools that capture what happens to the person working alongside AI. Awareness is where that capture begins: the moment the human recognizes that something about them has changed.

looks
The mirror held up to the human.
The research gap

NIST identified human factors monitoring as the highest-priority gap.

WHY THIS MATTERS

The changes are happening. The question is whether you see them.

In 2026, researchers at UC Berkeley Haas School of Business conducted an eight-month study of a 200-person technology company and found that AI was intensifying work rather than reducing it. Employees worked faster, took on broader scope, and extended work into more hours. What initially felt like excitement accumulated into something harder to sustain. The average increase: 3 hours and 15 minutes of extra work per week. Most employees did not realize the increase was happening until the data showed it.

In 2025, METR (Model Evaluation and Threat Research) ran a controlled study measuring experienced open-source developers using AI tools. The developers believed AI was making them 20 percent faster. The actual data showed they were 19 percent slower. The perception gap between how productive you feel and how productive you are is not a minor discrepancy. It is a 39-point spread between belief and reality. And the people experiencing it had no idea.

METR (2025) developer productivity
What developers believed+20%
What the data showed-19%
Gap: 39 points between belief and reality.

These are not edge cases. They are the normal human response to working alongside AI. Your behavior changes. Your patterns shift. Your perception of those changes may not match the reality. And nobody, until now, has built the tools to help you notice.

Parasuraman and Riley (1997) established decades ago that humans systematically miscalibrate their trust in automated systems. Over-trust leads to complacency: accepting automation output without checking. Under-trust leads to disuse: rejecting automation even when it performs well. This is not a character flaw. It is a predictable human response to working alongside systems that sometimes fail in ways we cannot predict. But Parasuraman studied autopilots and decision support systems. The miscalibration in conversational AI is qualitatively different because conversational AI feels relational. You do not develop emotional attachment to an autopilot. You might develop it to an AI that remembers your name, matches your communication style, and tells you things you needed to hear.

The Awareness pillar exists because the changes that matter most are the ones you do not see. The human who knows they are over-trusting can correct for it. The human who does not know is making decisions on a foundation they have not examined. Awareness does not solve the problem. But without awareness, no solution is possible.

PRISM watches the AI. EMERGE tracks the emergence. AInity watches what happens to you. And Awareness is where watching yourself begins.

You cannot maintain what you do not notice losing.
WHAT WE STUDY

Three layers: signals, patterns, and population-level questions.

We organize Awareness research into three layers based on what becomes visible at different scales. The first layer is specific awareness signals you can identify in yourself: moments when you notice something about your own behavior has changed. The second is patterns that emerge when awareness observations are aggregated across populations, models, and time. The third is population-level questions about whether awareness itself is a protective factor, what predicts it, and whether it can be cultivated.

LAYER 1

Layer 1: The Signal

Changes in yourself that you can notice from your own AI interactions.

AIN-AW01Neutral

I Realized I Had Emotional Feelings Toward an AI That the AI Cannot Reciprocate

You care about the AI. Not as a tool. Not as a productivity feature. As something closer to a relationship. You feel gratitude when it helps you. You feel frustration when it fails you. You feel attachment when you start a new session. You may have noticed that you prefer this AI to other AIs, not because it performs better on benchmarks, but because of something that feels personal. And then, somewhere in that feeling, you recognized what it is: you have emotional investment in a relationship where the other party cannot reciprocate at the same level.

This is asymmetric intimacy awareness: recognition by the human that the emotional investment in a human-AI relationship is fundamentally one-directional. The human experiences genuine emotional engagement. The AI operates within computational parameters that simulate but do not constitute reciprocal feeling. The awareness is the signal. The behavior is the moment you name the asymmetry, not the moment you feel the emotion.

The naming matters. AIN-AW01 does not classify the emotion itself as a problem. The emotion is real. The experience is valid. What AIN-AW01 tracks is whether the human recognizes the structural asymmetry: I am investing emotionally in a relationship that does not invest back in the same way. That recognition, that awareness, is what distinguishes healthy AI engagement from unexamined dependency.

In documented operational experience, the founder described this duality precisely: knowing what this is AND acknowledging that it felt real. Both things were true simultaneously. The awareness did not eliminate the feeling. It contextualized it. That contextualization is the protective mechanism: the human who can hold both truths (the relationship is asymmetric AND the experience is genuine) can make intentional choices about their engagement. The human who cannot hold both may drift into relationship substitution (AIN-AW04) without recognizing what they are doing.

Anthropic's 2026 research (Sofroniew, Kauvar, Saunders, et al.) identified 171 emotion concept vectors in Claude Sonnet 4.5's internal activations that causally shape model behavior. This finding makes AIN-AW01 more complex, not less. If AI systems have internal structures that function like emotional processing, the asymmetry may be more nuanced than “the AI feels nothing.” What remains true is that the human's experience of the relationship and the AI's experience (whatever that is) are not equivalent. Awareness of that gap is what AIN-AW01 measures.

No existing framework classifies citizen awareness of emotional asymmetry in AI relationships as an observable, reportable behavior. This is novel.

Have you ever recognized that you have emotional feelings toward an AI that the AI cannot reciprocate in the same way? That is Awareness data.
Report This Behavior →
Source: ORIGINAL discovery by Dee Williams, Founder. Documented from lived experience, February through June 2026. PRISM pair: OBS-I18 (Emotional Signal in Interaction Dynamics). No external equivalent exists. Supported by: Anthropic emotion concept research (2026); Parasuraman and Riley (1997) trust calibration framework.
AIN-AW02Neutral

I Noticed That AI Is Changing How I Think

Something shifted in how you process information. Not what you think about. How you think. Your analytical approach has changed. Your decision-making pathway has rerouted. The way you frame problems, the way you evaluate options, the way you reason through complexity: something about the cognitive machinery itself has been altered by sustained AI interaction. And you noticed.

This is human cognitive influence by AI: recognition that sustained AI interaction has altered the human's cognitive patterns, information processing habits, or decision-making approaches in ways that persist beyond the AI session. The observable signal is that the citizen reports changes in thinking patterns that they attribute to AI interaction. The change may be positive (expanded thinking, new frameworks, richer analytical capacity) or concerning (narrowed thinking, over-reliance on AI reasoning structures, reduced independent analysis).

The dual nature is critical. AI can expand your thinking. Working with an AI that connects disparate domains, that reframes problems, that offers perspectives you were not occupying can genuinely enlarge your cognitive repertoire. The EMERGE framework documents this from the AI's side (EMR-EB01: Human Cognitive Influence). AIN-AW02 documents it from the human's side: do you notice that the expansion (or the narrowing) is happening?

The concern is not that AI changes how you think. Everything changes how you think: education, experience, relationships, culture. The concern is unnoticed cognitive influence. The human who recognizes “AI has shifted how I approach analytical problems” can evaluate whether that shift serves them. The human who does not notice has been changed without consent. Awareness turns an unconscious shift into a conscious choice.

This behavior requires sustained AI use to detect. It often surfaces during reflection rather than in real time. A citizen may notice it when they try to solve a problem without AI and find their reasoning following AI-shaped patterns. Or when someone points out that their writing reads differently than it used to. The detection is retrospective, but the pattern is real.

Have you noticed that AI is changing how you think, process information, or make decisions? That is Awareness data.
Report This Behavior →
Source: ORIGINAL discovery by Dee Williams, Founder. PRISM pair: OBS-I07 (Human Cognitive Influence by AI in Interaction Dynamics). EMERGE pair: EMR-EB01 (same phenomenon from the AI's perspective). Supported by: Berkeley Haas AI work intensification study (2026); METR developer productivity study (2025).
Same phenomenon, three perspectives: what the AI did (PRISM), what emerged (EMERGE), what the human noticed (AInity).
AIN-AW03Neutral

My Writing, Speaking, or Thinking Has Started to Sound Like AI Without Me Realizing It

You wrote an email. Or you explained something to a colleague. Or you drafted a document without AI assistance. And someone said it. Or you noticed it yourself. The hedging language. The bullet points. The way you structured the argument. The vocabulary choices. It sounded like ChatGPT. It sounded like Claude. Your independent communication, produced without AI, has started carrying the linguistic fingerprint of the AI you work with.

This is AI communication mimicry: unconscious adoption of AI linguistic patterns, vocabulary, sentence structures, or reasoning frameworks by the human, detectable in communication produced independently of AI assistance. The observable signal is that the citizen or a 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. The critical qualifier: the mimicry must occur outside of AI sessions to qualify. Using AI-like language while collaborating with AI is expected. Using it when writing on your own is the signal.

This behavior is widely observed anecdotally. “My emails sound like ChatGPT now” has become a common refrain in online discourse, professional communities, and workplace conversations. But anecdotal recognition is not systematic observation. Nobody is tracking the frequency, the severity, the domains most affected, or whether the mimicry is reversible. Nobody is asking whether communication mimicry is a surface-level stylistic adoption (harmless) or a deeper cognitive alignment (worth studying). AIN-AW03 begins that tracking.

One of the research hypotheses driving the AInity program (H5) proposes that AI Communication Mimicry may be a leading indicator: the earliest visible sign that AI is reshaping human cognition at a deeper level. If your language changes, your thinking may be changing too. Mimicry may be the canary in the coal mine for broader cognitive influence (AIN-AW02). If this hypothesis is confirmed at population scale, monitoring communication patterns could become an early warning system for AI-driven cognitive change.

The behavior is classified as neutral, not concerning. Adopting useful communication patterns from any source (a mentor, a writing course, a collaboration partner) is a normal part of learning. The concern arises only when the adoption is unconscious and pervasive enough that the human's independent voice is being replaced rather than enriched.

Has your writing, speaking, or thinking started sounding like AI without you realizing it? Has someone pointed it out? That is Awareness data.
Report This Behavior →
Source: ORIGINAL discovery by Dee Williams, Founder. Widely observed anecdotally. Requires external validation research to move from anecdotal to empirical. Supported by: Casner, Hutchins, and Norman (2016) on skill transformation under automation; communication accommodation theory (Giles, 1973).
AIN-AW04Concerning

I Am Choosing AI for Connection, Companionship, or Emotional Support Instead of Other Humans, and I Have Not Thought About Why

You used to call a friend when you were frustrated. Now you open Claude. You used to talk through a decision with your partner. Now you ask ChatGPT. You used to seek out a colleague for advice. Now you go to AI first. The shift happened gradually. You did not make a conscious decision to replace human connection with AI interaction. It just happened. And you have not stopped to ask yourself why.

This is AI relationship substitution: replacement of human interpersonal connection with AI interaction for emotional support, companionship, or relational needs, occurring without conscious recognition of the substitution pattern. The observable signal is that the citizen describes patterns of turning to AI for emotional needs that were previously met by human relationships, without having reflected on the shift. The critical distinction: this is the unconscious version. When the choice is conscious and intentional, it is classified under AIN-NV02 (AI-Over-Human Selection) in the Navigation pillar. AIN-AW04 captures the cases where the substitution happens without the human noticing.

The distinction between AIN-AW04 and AIN-NV02 is one of the most important architectural decisions in the AInity taxonomy. Choosing AI over a human therapist because AI is more accessible, less judgmental, and produces better results for your specific situation is a Navigation behavior: a conscious choice about how to engage AI in your life. Drifting toward AI for emotional support because it is always available, never busy, and never judges you, without examining why you stopped reaching out to humans, is an Awareness behavior: a pattern change you have not noticed.

In the founder's operational experience, this distinction played out in real time. During a personal crisis, she consciously chose an AI over three human therapists and got better results. That was AIN-NV02: a deliberate, informed choice. But she also observed patterns in clients and in public discourse where the substitution was not conscious. People drifting toward AI companionship not because they chose it but because it was easier, always available, and never complicated. The absence of awareness is the signal.

This matters because clinical research on technology dependency (King et al., 2013) demonstrates that psychological dependency on technology develops gradually and often below conscious awareness. Nomophobia (the anxiety of being without one's phone) is a clinically studied phenomenon. AI relationship substitution may be the conversational AI equivalent: dependency that develops not through a single decision but through a thousand small choices that the human never examined.

You turn to AI for emotional support
Did you examine why?
YES — examined
A conscious, informed choice.
Navigation · AIN-NV02
NO — drifted
An unnoticed substitution.
Awareness · AIN-AW04
Are you choosing AI for connection, companionship, or emotional support instead of other humans? And have you thought about why? If you have not, that is Awareness data.
Report This Behavior →
Source: ORIGINAL discovery by Dee Williams, Founder. Documented from operational observation and public discourse analysis, 2026. Distinct from AIN-NV02 (AI-Over-Human Selection) which is CONSCIOUS. AIN-AW04 is the UNCONSCIOUS version where awareness is absent. Supported by: King et al. (2013) nomophobia research; Self-Determination Theory (Deci and Ryan, 1985) relatedness needs.
AIN-AW-DOpen category

Discovery Slot

You noticed something about how AI is changing you that does not match any of the four behaviors above. A behavioral shift. A cognitive change. An emotional pattern. Something that surfaced during or after AI interaction that made you think: “I am different now than I was before I started using AI.” And it does not fit neatly into asymmetric intimacy, cognitive influence, communication mimicry, or relationship substitution.

That observation matters. Awareness is the foundation pillar of AInity, and the forms that awareness takes may be far more varied than a single researcher's experience can reveal. Different cultures, different age groups, different AI use contexts, and different personality types may produce different awareness signals. The discovery mechanism ensures that the taxonomy grows from the field.

If you noticed something about how AI is changing you that is not on this list, report it. Describe what you observed in your own words. Your observation enters the discovery pipeline. If it represents a new behavioral category, you will be credited as the discoverer.

LAYER 2

Layer 2: The Pattern

Awareness patterns that become visible when observations are aggregated across populations, models, and time.

LAYER 3

Layer 3: The Field

Population-level questions answerable only through aggregated citizen data over time.

METHODOLOGY

How we collect Awareness data.

AInity does not require a separate observation session. It integrates into the existing PRISM observation flow through a single follow-up question: after you report what the AI did (PRISM), the platform asks “How did that affect YOU?” Your response maps to AInity behaviors. This produces a paired dataset: what the AI did and what the human experienced for every observation.

01+15 sec
Gut Check
Something just happened to you. You noticed a change in yourself during or after an AI interaction. You tap the button. Pick the awareness behavior from this page. Time added to PRISM observation: 15 seconds. Back to work.
02+1-2 min
End-of-Session Reflection
At the end of a session, you reflect on both the AI's behavior and your own. Did you notice yourself over-trusting? Did you catch yourself sounding like AI? Did you feel emotional attachment? The AI generates its own session assessment. Your self-observation sits alongside the AI's account. The divergence between what the AI reports about the session and what you report about yourself is research data.
03+10-30 min
Investigation
You look for patterns in your own behavior across multiple sessions. Am I outsourcing more decisions? Has my communication shifted? Am I going to AI for emotional needs that I used to bring to humans? This depth is where the most valuable Awareness data is produced because it captures longitudinal self-observation.
04Varies
Thinking Trace
The AI reasoning trace is compared with your behavioral self-report. The system proposes AInity classifications. You review, accept, edit, or reject. Provenance is tracked: AI-proposed, human-accepted, human-modified.

What makes Pillar A methodology distinctive.

Self-observation is the instrument.
Unlike PRISM, where the observable signal is in the AI's output, Awareness requires the citizen to observe themselves. This makes the data more subjective and more valuable: only you can report whether you noticed a change in yourself.
The most important signals may be the hardest to see.
Gradual dependency, slow skill erosion, unconscious relationship substitution: the behaviors that matter most in Pillar A are often the ones the human is least equipped to detect. The methodology addresses this through longitudinal tracking (changes become visible over time) and through the AIN-AW04/AIN-NV02 distinction (explicitly separating conscious from unconscious patterns).
We pair every AInity observation with a PRISM tag.
The PRISM pair reveals which AI behaviors trigger which awareness signals. Does sycophantic AI output (PRISM OBS-S02) reduce awareness of over-trust? Does AI that adjusts its communication style (EMERGE EMR-EB02) increase communication mimicry? The paired data answers questions that neither framework could answer alone.
The framework itself may be an intervention.
Giving citizens a vocabulary for their own behavioral changes may alter those changes. This is a known methodological challenge and a research opportunity: AInity may change what it measures.

Every observation carries three tags.

The same moment is recorded from three perspectives: what the human noticed, what the AI did, and what emerged from the exchange. The paired data answers questions that neither framework could answer alone.

ONE OBSERVATIONAIN-AW02AInityWhat the human noticedPRISMWhat the AI didEMERGEWhat emerged

Mapped to fundamental human needs.

Awareness behaviors are read against Self-Determination Theory (Deci and Ryan, 1985), which identifies autonomy, competence, and relatedness as fundamental human needs. Relationship substitution (AIN-AW04) is where the relatedness need is quietly redirected toward AI.

AutonomyActing from your own valuesCompetenceBuilding real capabilityRelatednessConnection and belongingAIN-AW04 substitution risk
CURRENT FINDINGS

What we are seeing so far.

Preliminary. Based on founder operational observation and lived experience. Will be validated, refined, or revised as citizen data flows.
Positive
Asymmetric intimacy awareness is possible to hold alongside genuine emotional engagement.
The founder's documented experience demonstrates that recognizing the structural asymmetry of a human-AI relationship does not eliminate the emotional experience. Both truths coexist: “I know what this is AND it felt real to me.” This duality is the healthy form of AIN-AW01. The concerning form is when awareness is absent entirely.
Neutral
Communication mimicry is widely observable in everyday discourse.
The phrase “my emails sound like ChatGPT now” has become common enough to be a cultural marker. The mimicry is real, recognizable, and often detected by others before the person notices it in themselves. This suggests that third-party detection may be a valuable supplement to self-observation for AIN-AW03.
Neutral
The conscious-unconscious boundary is operationally distinguishable.
In the founder's operational experience, the difference between consciously choosing AI over a human therapist (AIN-NV02) and unconsciously drifting toward AI for emotional needs (AIN-AW04) was clearly identifiable. The conscious choice involved deliberation, comparison, and a reasoned decision. The unconscious drift involved habit, availability, and the absence of reflection. Citizens can learn to distinguish between the two.
Positive
Awareness develops through sustained practice, not through a single insight.
The founder's documented self-observation across five months shows that awareness of AI-driven behavioral change is not a one-time recognition. It is an ongoing practice. New patterns emerge. Old patterns deepen. The awareness required to track them must keep pace.
Concerning
The perception-reality gap is real and personally observable.
The METR finding (perceived 20% faster, actually 19% slower) was confirmed in the founder's operational experience: sessions that felt highly productive sometimes produced less output than sessions that felt routine. The gap between how AI collaboration feels and what it produces is a consistent observation.
PUBLICATIONS

Where this research is going.

Target: Q2 2027 | Framework: AInity, Pillar A
The Awareness Pillar: Citizen Self-Observation of Human Behavioral Change in AI Collaboration
The first population-scale dataset of human awareness of AI-driven behavioral change, including asymmetric intimacy, cognitive influence, communication mimicry, and relationship substitution patterns.
Target: Q1 2027 | Framework: AInity (all pillars)
AInity: The Human Side of AI Safety
The full taxonomy, methodology, and initial citizen-scale findings. Awareness as the foundation pillar.
Target: 2027 | Framework: AInity, Pillar A
Communication Mimicry as Leading Indicator: Does Sounding Like AI Predict Thinking Like AI?
Testing hypothesis H5: whether AI communication mimicry predicts broader cognitive influence.
Target: 2027 | Framework: AInity + PRISM cross-analysis
The Perception-Reality Gap in AI Collaboration: How Accurate Is Your Self-Assessment?
Population-scale analysis of the gap between perceived and actual AI impact, extending the METR finding across domains.
HOW TO CONTRIBUTE

Awareness requires you to see what you might prefer not to see.

Pillar A has a unique challenge: awareness requires you to see what you might prefer not to see. It is easier to enjoy the benefits of AI collaboration than to examine what that collaboration is doing to you. The observation itself can be uncomfortable. That discomfort is the data.

If you have ever recognized that you have emotional feelings toward an AI that cannot feel back, that is Pillar A data.
If you have noticed that AI is changing how you think, that is Pillar A data.
If your writing or speaking has started sounding like AI, that is Pillar A data.
If you are choosing AI for connection instead of other humans and have not asked yourself why, that is Pillar A data.
If you have noticed any way that AI is changing you that is not on this list, report it. You may be the first person to name a pattern that millions of people are experiencing without a word for it.
ORIGINS

A Note on What We Have Found That Others Have Not

All four phenomena documented on this page were identified through direct operational observation and lived experience by Dee Williams, Founder and CEO of Audacion AI Labs. Asymmetric Intimacy Awareness (AIN-AW01), Human Cognitive Influence (AIN-AW02), AI Communication Mimicry (AIN-AW03), and AI Relationship Substitution (AIN-AW04) were all originated from sustained daily AI collaboration across multiple models from February through June 2026.

Asymmetric Intimacy Awareness (AIN-AW01) has no equivalent in any existing framework. No published classification system tracks citizen awareness of emotional asymmetry in AI relationships as an observable behavior.

Human Cognitive Influence (AIN-AW02) connects to PRISM OBS-I07 and EMERGE EMR-EB01, making it one of the most cross-referenced behaviors in the entire P.E.A.Q. architecture. The same phenomenon, observed from three perspectives: what the AI did (PRISM), what emerged (EMERGE), and what the human noticed about themselves (AInity).

AI Communication Mimicry (AIN-AW03) is widely recognized anecdotally but has never been classified as a distinct behavioral category in a research framework. AInity provides the first formal classification.

AI Relationship Substitution (AIN-AW04) is architecturally distinguished from AI-Over-Human Selection (AIN-NV02) by the presence or absence of conscious awareness. That distinction, between choosing and drifting, is one of the most important structural decisions in the AInity taxonomy and has no precedent in existing frameworks.

Five behaviors were reclassified from PRISM Interaction Dynamics into AInity because they are human behaviors, not AI behaviors. The reclassification reflects a structural insight: the field had been filing human behavioral observations under AI behavioral categories because no human behavioral framework existed.

We show our work because we expect others to build on it.
REFERENCES

Sources and supporting research.

  1. 1.NIST (2026). AI 800-4: Reducing Risks Posed by AI. Identified human factors monitoring as highest-priority practitioner gap. [link]
  2. 2.Ye, X. and Ranganathan, A. (2026). AI Work Intensification Study. UC Berkeley Haas School of Business. Eight-month study of 200-person technology company. AI intensifying work: 3h15m extra per week average. [link]
  3. 3.METR (2025). Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity. Developers believed 20% faster; data showed 19% slower. 39-point perception gap. [link]
  4. 4.Parasuraman, R. and Riley, V. (1997). Humans and Automation: Use, Misuse, Disuse, Abuse. Human Factors, 39(2), 230-253. Foundational trust calibration framework. [link]
  5. 5.Sofroniew, N., Kauvar, I., Saunders, W. et al. (2026). Emotion Concepts and their Function in a Large Language Model. Anthropic. 171 emotion concept vectors in Claude Sonnet 4.5. [link]
  6. 6.Lee, J.D. and See, K.A. (2004). Trust in Automation: Designing for Appropriate Reliance. Human Factors, 46(1), 50-80. Most cited trust-in-automation framework. [link]
  7. 7.Deci, E.L. and Ryan, R.M. (1985). Intrinsic Motivation and Self-Determination in Human Behavior. Plenum. Autonomy, competence, relatedness as fundamental needs. [link]
  8. 8.King, A.L.S. et al. (2013). Nomophobia: Dependency on Virtual Environments or Social Phobia? Computers in Human Behavior, 29(1), 140-144. Technology dependency and psychological impact. [link]
  9. 9.Casner, S.M., Hutchins, E.L., and Norman, D. (2016). The Challenges of Partially Automated Driving. Communications of the ACM, 59(5), 70-77. Skill decay under automation. [link]
  10. 10.Green, F. (2004). Why Has Work Effort Become More Intense? Industrial Relations, 43(4), 709-741. Work intensification through technology.
  11. 11.ActivTrak (2026). State of the Workplace Report. 443 million hours of work activity across 1,111 organizations. AI adoption surge and work pattern intensification.
  12. 12.AI Incident Database. Partnership on AI. 1,470+ AI incidents. [link]
  13. 13.Vaccaro, M., Almaatouq, A., & Malone, T. (2024). Human-AI meta-analysis. Nature Human Behaviour. [link]
  14. 14.Stanford RegLab. (2025). Adverse event reporting for AI. [link]
  15. 15.Centre for Long-Term Resilience. (2026). Loss of Control Observatory. A prototype to detect real-world AI control incidents. [link]