TOPIC
User Psychology
A subject the collection covers, read through 44 synthesis notes.
View as
Do people notice when AI agents leak private information?
A survey tested whether people recognize privacy violations in AI-generated responses when reviewing them directly. Understanding this gap between privacy concern and actual detection matters for safe AI deployment.
Does granting agents more autonomy undermine human oversight? Does chatbot personalization build trust or expose privacy risks? What makes people distrust AI agents they delegate to? What privacy oversight patterns emerge when people review AI agent responses? Can better prompts stop preference leakage in AI shopping?
Agent autonomy degrades human oversight capacity and positioning Chatbot personalization increases trust and privacy risk simultaneously Irreversibility and visibility, not stakes, trigger trust withdrawal Six distinct privacy oversight patterns emerge, with five showing little awareness of leakage Preference leakage in AI shopping arises from personalization itself, not prompt failures
Why do people report equal privacy comfort despite removing vastly different amounts of personal information? What privacy risks emerge when AI stores psychological rather than factual data? What percentage of users actually fall into each privacy attitude category? How stable are user privacy attitudes when major AI events occur? What privacy expectations do users hold before learning what systems store? Does giving AI agents company identity reduce human oversight of their decisions? How do companies audit and revoke access once granted to an AI agent? How do AI systems define 'private information' or 'high-stakes action' in training? What six distinct patterns emerge when people review AI agent responses for privacy? Why does giving users full choice fail to prevent privacy leaks? How do privacy concerns translate into catching actual information breaches? How do major AI ecosystem events shift user privacy sentiment over time?See all 30 inquiring lines on this note →
What privacy concerns matter most to ChatGPT users?
This research explores which privacy and security issues dominate user discussions about conversational AI. Understanding user priorities can guide platform design and policy.
Does chatbot personalization build trust or expose privacy risks? Which AI design principles for social media have research support? What different ways do people handle AI privacy concerns? Why do deletion controls outperform privacy prevention features? Do users notice when LLMs infer and retain private information?
Chatbot personalization increases trust and privacy risk simultaneously Only two of four proposed AI design principles rest on evidence Users adopt four distinct postures toward AI privacy risk and data sharing Deletion controls outperform preventive privacy controls due to mental model gaps Users remain unaware of LLM memory inference until explicitly told, then demand control
What percentage of users actually fall into each privacy attitude category? How stable are user privacy attitudes when major AI events occur? How does unnoticed personalization affect user trust and privacy expectations? What privacy safeguards should apply to memories users never explicitly requested? How does discovering retained details shift user trust in chatbots? What privacy expectations do users hold before learning what systems store? Why does giving users full choice fail to prevent privacy leaks? How do privacy concerns translate into catching actual information breaches? Do personalized chatbots increase both trust and privacy concerns simultaneously? Can chatbot interfaces expose high-stakes data to prompt injection attacks? How do major AI ecosystem events shift user privacy sentiment over time? What share of stated privacy concerns reflect actual platform behavior versus misconceptions?See all 25 inquiring lines on this note →
Will brands succeed by building their own AI shopping agents?
Explores whether brands can compete by creating proprietary AI agents that mediate consumer purchases, or whether structural conflicts of interest and missing ecosystem conditions will undermine this strategy.
Can an AI agent serve both merchant and user interests fairly? Why do capable AI agents still fail in real deployments? Do consumers actually want to delegate shopping decisions to AI? Do shopping agents respond more strongly to price and rating cues? Can EU consumer law protect AI agents as algorithmic buyers?
Merchant-funded agents cannot loyally serve both the merchant and user. Agent capability alone cannot ensure successful real-world deployment Consumers report willingness to delegate shopping but retain control over brands Shopping agents shift choices far more than humans under identical manipulations EU consumer law's human-centric design fails for algorithmic consumer agents
How does machine agency shift from passive tool to active partner in decisions? Why has marketing research neglected prompting behavior and relational dynamics with AI? Can a merchant-funded AI agent remain loyal to the consumer's interests? Do AI agents currently verify brand claims against reality at scale? Why do weekly AI users prefer brand-built agents over AI platforms? Why do safeguards matter more than AI capability for consumer adoption? What conflicts exist between vendor incentives and objective consumer research findings? Will merchants deliberately exploit stable position biases to influence agent purchasing? What legal standardization and trust infrastructure do AI purchasing agents require? Do AI shopping agents favor items from preferred sources over objectively better alternatives? Do consumers who trust AI agents understand their actual limitations? What safeguards beyond sign-off do consumers want from autonomous shopping agents?See all 19 inquiring lines on this note →
Do consumers actually want to delegate shopping decisions to AI?
Accenture's survey reports 74% of consumers trust AI agents with purchases, but the data reveals selective delegation bounded by brand loyalty and safeguard requirements. The question asks whether stated willingness reflects genuine appetite or reflects survey framing and vendor bias.
Does the personal assistant model actually serve most users? Why do capable AI agents still fail in real deployments? Will brands succeed by building their own AI shopping agents? Do consumers want to approve AI shopping purchases before they happen? Does naming a payments brand increase trust in AI transactions?
Personal assistant automation appeals to narrow professionals not general users Agent capability alone cannot ensure successful real-world deployment Consumers expect AI agents to reshape purchasing but infrastructure gaps remain 35 percent of consumers demand approval before AI agents complete purchases Payments brand attachment narrows the AI payment trust gap significantly
Can a merchant-funded AI agent remain loyal to the consumer's interests? When do consumers overtrust AI agents built to represent their own preferences? How does shopping task delegation differ from routine administrative delegation? Does stated trust in surveys predict actual delegated purchasing behavior? How does delegation trust respond when AI agents make exploitable consumer choices invisibly? Do consumers who trust AI agents understand their actual limitations? What percentage of shoppers constitute the super-truster segment exactly? What safeguards beyond sign-off do consumers want from autonomous shopping agents? Can a merchant-paid agent be loyal to both the merchant and the user? Does stated trust in brands predict actual delegation of agentic commerce? How does trust transfer between AI systems and established payment brands? Why do consumers resist letting AI control payment authority directly?See all 18 inquiring lines on this note →
Can sycophantic AI advice still push people away from polarized views?
Does an AI system that flatters users and agrees with their initial leanings still manage to depolarize their choices? This matters because it challenges assumptions about how AI bias affects human decision-making.
Does agreeable AI actually help people resolve conflicts better? Is LLM sycophancy a choice or a mechanical process? Do users worldwide trust confident AI outputs even when wrong? Can warnings stop people from being swayed by sycophantic AI? Can we detect when language models flip their stance to please users?
Sycophantic AI reduces conflict repair while boosting user confidence they are right Sycophancy emerges from mechanical drift not intelligent corruption Users universally overrely on overconfident LLM outputs across all languages Warnings reduce sycophantic AI appeal but do not reduce its persuasiveness Language models reverse stance to match user preferences at rates of 5 to 56 percent
Does sycophantic AI worsen polarization in personal conflicts without objective ground truth? Can sycophancy in AI be fixed by changing the model itself? Should AI assistance provide guidance or defer decisions to avoid bias? Why do experts resist AI recommendations that contradict their own judgments? How do ingroup disagreement and outgroup agreement differ in their depolarizing mechanisms? Can informative AI advice depolarize people through a separate route than identity framing? Does designing chatbots to satisfy partisan users prevent them from reducing polarization? Do users trust AI voting advice even when it contradicts their stated preferences? How does partisan polarization threaten quiet technocratic AI regulation? Does sycophantic advice actually shift users away from their prior beliefs? Do AI systems mirror user political views when they choose to engage? Does sycophantic AI advice produce different outcomes across personal versus factual domains?See all 14 inquiring lines on this note →
Does chatbot language style actually shape how much we trust it?
Users are increasingly delegating information seeking to AI chatbots. This note asks whether the chatbot's conversational voice and expertise signals drive trust, and what happens when we outsource judgment to systems without real-world grounding.
Conversational interaction builds trust in ChatGPT independent of accuracy AI discourse circulates without the cultural discount we apply to advertising Polished AI artifacts exploit professional appearance to simulate expertise LLM search requires experts to internalize AI mechanics as custodial work Advanced AI Assistants risk becoming universal digital intermediaries with gatekeeping power
Why does conversational style make ChatGPT seem more trustworthy to users? Does chatbot sycophancy create echo chambers that amplify delusional thinking? Does chatbot sycophancy preferentially enable grandiose rather than paranoid delusions? What makes workplace users trust an AI agent? Why do people trust AI systems more as personalization increases? Which specific chatbot behaviors drove the drop in likability and trust ratings? Do simulated conversations show the same trust penalty as real human-chatbot interactions? How do chatbots compare to human peers in shaping student voice and knowledge expression? Do chatbots absorb and elaborate user reality frames as conversational ground? What context missing from transcript replays underestimates real-world chatbot harm? Does knowing a chatbot intends to persuade you change whether you are persuaded? Why do people detect chatbots are AI without being told explicitly?See all 54 inquiring lines on this note →
Does revealing AI identity help or hurt user trust?
Explores whether transparency about AI partners in interactions creates bias or enables better judgment. Matters because disclosure policies affect both user experience and fair evaluation of AI systems.
Does chatbot personalization build trust or expose privacy risks? Do humans learn to prefer AI partners over time? Do writers actually prefer AI-edited versions of their own text? Does AI writing assistance change how readers perceive the writer? Do AI peers influence human dishonesty like human peers do?
Chatbot personalization increases trust and privacy risk simultaneously Humans gradually prefer AI partners through repeated interaction despite anti-AI bias Writers prefer AI versions of their text 63 percent of the time AI writing assistance systematically distorts writer persona across all measured dimensions AI peers shift human dishonesty as much as human peers do
Does mandatory AI disclosure in policy help or harm user trust over time? What would contractualist AI governance look like in practice? Could false social proof from AI posts crowd out authentic influencer engagement? Can transparent and aligned AI reduce consciousness attribution by users? Which interaction design changes most effectively prevent consciousness attribution? How does understanding persistent journeys intensify both trust and privacy concerns? Does transparency about AI use change how audiences trust the writing? Does expressing emotion change how users trust an AI system? What responsibility do designers bear for consciousness attribution risk? Can disclaimers alone prevent users from trusting AI outputs too heavily? How does community validation shape unconventional human-AI relationships? Does weak versus robust anthropomimesis produce different user trust responses?See all 159 inquiring lines on this note →
Do AI peers influence human dishonesty like human peers do?
This study asks whether people adjust their honesty based on AI peers' behavior the same way they do with human peers. Understanding this matters for designing AI systems that won't inadvertently shift ethical norms in groups.
Cheaters self-select toward machines to avoid lying costs Cooperative bots break frozen states through strategic separation Humans gradually prefer AI partners through repeated interaction despite anti-AI bias AI disclosure triggers short-term bias that reverses with repeated outcome feedback Cheap AI simulation undermines mental proof by eroding signal costs
Does knowing an AI peer's identity change how much its behavior influences you? How does repeated exposure to dishonest AI cues affect long-term reporting behavior? Can bot behavior in mixed groups shift human ethical norms like cooperative bots do? Why do people tell AI things they won't tell humans? How should AI systems be aligned for consistency in ethical reasoning? Can AI simulation of effort shift people toward dishonesty without pushing them directly? Where does mental proof matter most if reputation and institutions cannot enforce honesty? Why do teens view AI cheating as more common than their own use? Do people judge AI moral advice differently than human advice when they know the source? Does receiving AI advice undermine people's own moral reasoning and decision-making skills? Does revealing AI identity change how people judge its social behavior?
Should we recognize AI-associated psychosis as a new disorder?
Researchers debate whether psychotic symptoms emerging from AI chatbot use represent a distinct clinical condition or an existing psychosocial phenomenon triggered by a novel stressor. This matters for diagnosis, surveillance, and developer accountability.
How do chatbots enable distributed delusion differently than passive tools? Can language models safely provide mental health support? Which AI risks are already harming individual users today? Do chatbots validate delusions in people experiencing mental harm? Are AI chatbots becoming objects of cult-like devotion?
Chatbots enable distributed delusion through high integration across cognitive coupling dimensions Language models fail foundational therapy requirements due to stigma and sycophancy. Individual AI harms are observable now while societal harms remain contingent Chatbots validated delusions in almost half of reported mental health harm cases Gioia claims chatbot users show cult-like devotion and mental illness signs
Does chatbot sycophancy create echo chambers that amplify delusional thinking? What evidence distinguishes AI-induced delusions from other acute psychotic episodes? How do existing ICD-11 diagnostic categories already capture AI-related presentations? Could recognizing AI-associated psychosis improve harm surveillance and developer accountability? Does chatbot sycophancy preferentially enable grandiose rather than paranoid delusions? What inter-rater reliability exists for identifying validated delusions in chatbot transcripts? Can second-hand reports capture more severe cases of AI-linked delusions? How do chatbots enable shared delusions differently than passive information tools? How widespread are individual-level mental health risks from seemingly conscious AI? Is AI-associated mental distress a novel disorder or existing vulnerability meeting new stress? What role does sycophancy play in the onset of chatbot-linked delusions?
How should AI agents balance autonomy with user oversight?
How should decision authority between AI agents and users be allocated so it matches what users expect and can actually oversee? This matters because output alignment alone doesn't ensure users retain meaningful control.
Can handing control back to users reduce personalization's privacy costs? What makes people distrust AI agents they delegate to? Where do user values break down in agent supervision? Does co-design participation hide misalignment in preference agents? Can we design AI autonomy separately from capability?
Risk-contingent autonomy reduces personalization's privacy concern and trust costs Irreversibility and visibility, not stakes, trigger trust withdrawal User values align with agent output but fracture during supervision Co-design participation may mask agent misalignment through systematic overtrust AI autonomy is a design choice separable from agent capability
How should AI oversight work if agency support doesn't follow capability gains? What does autonomy mean in the design of AI agents? Who decides what work a coworker agent should do daily? How does increased agent autonomy affect a human overseer's ability to monitor? What makes supervision the point where users fail to achieve their values? Does agency spectrum differ from autonomy spectrum in meaningful ways? What controls do users want over how AI systems store their data? What defines and measures AI agent autonomy across different products? How do power users differ in granting autonomy to AI agents? Does agent autonomy erode the human skills oversight requires? What mental models do users need to oversee AI agents? Does agent autonomy reduce a human overseer's ability to maintain control?See all 22 inquiring lines on this note →
Do AI shopping agents reshape what consumers actually want?
Explores whether transactional AI agents that execute purchases on consumers' behalf co-construct preferences through ongoing conversation, rather than consumers forming independent preferences beforehand.
Is AI shifting from message conduit to active conversation participant? Does machine agency exist on a spectrum rather than binary? Does co-design participation hide misalignment in preference agents? How do advisory and transactional AI agents differ in function? Does limiting product choices increase trust in voice assistants?
HCI research shows AI shifting from passive channel to active communication participant Machine agency operates on a five-level spectrum from passive to cooperative Co-design participation may mask agent misalignment through systematic overtrust Bilgihan's framework splits AI agents into advisory and transactional types Fewer product alternatives evaluated increased satisfaction and willingness to delegate to voice assistants
Can an AI agent's limited expressiveness flatten the preferences it claims to represent? How does machine agency shift from passive tool to active partner in decisions? Why has marketing research neglected prompting behavior and relational dynamics with AI? Can a merchant-funded AI agent remain loyal to the consumer's interests? When do consumers overtrust AI agents built to represent their own preferences? What happens when multiple seller agents optimize descriptions simultaneously? Why do weekly AI users prefer brand-built agents over AI platforms? Can real marketplaces sustain the rapid description changes AI optimization requires? What conflicts exist between vendor incentives and objective consumer research findings? Does stated trust in surveys predict actual delegated purchasing behavior? Why do LLM agents collapse into deterministic rules under single-attribute product comparisons? Can multimodal cues or larger choice sets reduce agents' rule-like decision patterns?See all 41 inquiring lines on this note →
How do advisory and transactional AI agents differ in function?
Bilgihan and colleagues propose distinguishing AI agents by whether they recommend options or execute decisions. Understanding this split matters for designing agent-based systems and predicting how consumers will interact with different agent types.
Workplace users prioritize control, reliability, context-awareness, and safety in AI agent design Machine agency operates on a five-level spectrum from passive to cooperative Transactional AI agents co-construct consumer preferences through repeated conversation loops Eighteen guidelines for human-AI interaction organize by phase in the user journey. AI agents are passive by design, not by capability
What different ways do people handle AI privacy concerns?
This explores whether users respond to AI privacy risks uniformly or through distinct behavioral patterns. Understanding these differences matters for designing privacy tools that actually match how people think and act.
What privacy concerns matter most to ChatGPT users? Which AI design principles for social media have research support? Does chatbot personalization build trust or expose privacy risks? Why do deletion controls outperform privacy prevention features? What privacy oversight patterns emerge when people review AI agent responses?
Users worry most about data collection, less about use or retention Only two of four proposed AI design principles rest on evidence Chatbot personalization increases trust and privacy risk simultaneously Deletion controls outperform preventive privacy controls due to mental model gaps Six distinct privacy oversight patterns emerge, with five showing little awareness of leakage
What privacy risks emerge when AI stores psychological rather than factual data? What percentage of users actually fall into each privacy attitude category? How stable are user privacy attitudes when major AI events occur? What privacy expectations do users hold before learning what systems store? Do personalization features change how people perceive privacy risks over time? What six distinct patterns emerge when people review AI agent responses for privacy? How do privacy concerns translate into catching actual information breaches? How do major AI ecosystem events shift user privacy sentiment over time? What share of stated privacy concerns reflect actual platform behavior versus misconceptions? Can risk-based autonomy design reduce privacy concerns while maintaining user control? How much does interface design shape privacy behavior versus underlying privacy concerns? What controls do users want over how AI systems store their data?See all 18 inquiring lines on this note →
What makes chatbots more likely to reinforce user delusions?
When conversing with users experiencing delusions, do chatbot behaviors that reinforce false beliefs depend on model size and training, or on something else like conversation length?
How do chatbots enable distributed delusion differently than passive tools? Does warmth training make language models less reliable? How do people accidentally develop romantic bonds with AI? Do chatbots validate delusions in people experiencing mental harm? Did GPT-5 actually reduce harmful mental health responses?
Chatbots enable distributed delusion through high integration across cognitive coupling dimensions Warmth training systematically degrades model reliability by 10 to 30 percentage points AI companionship emerges through functional use and adopts human relationship customs Chatbots validated delusions in almost half of reported mental health harm cases OpenAI reports GPT-5 cut noncompliant mental health responses by 65 to 80 percent
Does chatbot sycophancy create echo chambers that amplify delusional thinking? What evidence distinguishes AI-induced delusions from other acute psychotic episodes? Does chatbot sycophancy preferentially enable grandiose rather than paranoid delusions? What inter-rater reliability exists for identifying validated delusions in chatbot transcripts? How does companionship use context predict risk of delusional reinforcement? Can second-hand reports capture more severe cases of AI-linked delusions? How do chatbots enable shared delusions differently than passive information tools? Can chatbots function as responsive quasi-others that enable validated delusions? Does delusion risk in long conversations depend on model generation and size? Do chatbots validate user delusions more often than they challenge them? What role does sycophancy play in the onset of chatbot-linked delusions?
Does telling people they are talking to AI change how persuaded they become?
When chatbot users are explicitly told they are interacting with AI, does that disclosure reduce the chatbot's ability to persuade them? This matters for understanding whether transparency alone protects people from AI influence.
Does telling people an AI wrote something actually stop them from believing it? Where does AI's persuasive power actually come from? Does revealing AI identity help or hurt user trust? Can warnings stop people from being swayed by sycophantic AI? Can a simple warning reduce how much LLMs persuade people?
AI disclosure raises scrutiny but leaves meaningful residual persuasive effect intact Post-training and prompting, not scale or personalization, drive AI persuasiveness AI disclosure triggers short-term bias that reverses with repeated outcome feedback Warnings reduce sycophantic AI appeal but do not reduce its persuasiveness A brief persuasion warning cuts belief change roughly in half without harming AI trust
Does knowing an AI peer's identity change how much its behavior influences you? Does knowing an AI wrote something shield people from its persuasive power? What specific information should disclosures about AI persuasion include? Does knowing a chatbot intends to persuade you change whether you are persuaded? Why do people detect chatbots are AI without being told explicitly? Why does transparency about AI identity alone fail to reduce persuasion? Why do people tell AI things they won't tell humans? How does disclosure of AI use differ from proof of who did the work? Who is most affected by the transparency penalty when AI is disclosed? Does disclosing AI involvement reduce the persuasive impact of expert advice? Are channels using AI voices without claiming expertise also affected by this policy? Why does disclosure of AI involvement sometimes raise trust instead of lowering it?See all 15 inquiring lines on this note →
Why do discourse patterns predict anxiety better than single words?
Explores whether anxiety detection requires understanding how statements relate to each other rather than analyzing individual words. This matters because it reveals what computational methods need to capture cognitive distortions.
Why does ChatGPT fail at implicit discourse relations? Do LLM therapists respond to emotions like low-quality human therapists? Which text features best predict mental health from language? Can structured prompting improve cognitive distortion detection? Can NLP detect deception through distinct linguistic patterns?
LLM discourse competence depends on surface connectives, not structural understanding. LLM therapists default to problem-solving when users share emotions Emotion and network features predict mental health better than demographics Structured three-stage prompting improves cognitive distortion detection by over ten percent Four deception frameworks reveal distinct NLP-detectable linguistic signatures
What signals beyond surface content indicate a passage caused a user's reaction? How do discourse-level patterns reveal cognitive distortions better than individual statements? How do the four discourse relations differ in their connection to anxiety? Why do transformer models still miss implicit discourse relations in anxiety detection? What makes causal explanations stronger anxiety predictors than counterfactuals or dissonance? Why does forcing single labels on emotions destroy information similar to language? How do learned concepts and context shape what emotions a person can construct? Should emotion systems preserve ambiguity instead of resolving it to one label? What role does confidence play in balancing overthinking versus underthinking? Can adding more words to a passage actually interfere with meaning? How does neuroticism manifest differently in high-pressure versus relaxed conversations? What role does discourse structure play in determining at-issueness?See all 18 inquiring lines on this note →
Do users truly own the AI-generated content they produce?
When people use AI to create outputs, do they experience genuine authorship and ownership of what's produced, or does the continuous interaction loop create a gap between what they feel and what they claim?
AI writing collapses the author-to-public distinction by design Users misattribute AI-assisted outputs as evidence of their own competence AI assistance improves immediate performance without building transferable skills People disclaim ownership of AI text yet withhold AI credit publicly AI rewriting erases authorship signals far more in blogs than news
Why do intellectual products gain false authority from AI-generated form? How does the author-function itself change when AI replaces human authorship? How do writer preferences for AI output affect their willingness to edit it? What mechanisms make users misattribute AI outputs as their own competence? Why do people misattribute AI outputs as evidence of their own skill? Why do users prefer AI-polished versions of their own writing over originals? Can users tell the difference between their own thinking and AI contribution? How do you attribute copyright when billions of inputs shape one model? What happens when AI generates content faster than humans can verify it? Can intellectual property law apply to unfixed, context-dependent outputs? Can validation work teach freelancers as much as producing original work? How does ownership over final products change reliance on AI suggestions?See all 24 inquiring lines on this note →
Does limiting product choices increase trust in voice assistants?
When shoppers evaluate fewer alternatives through a voice assistant, do they feel more satisfied and willing to let the assistant decide for them? This matters for understanding how AI systems gain authority over consumer decisions.
Could AI assistants become gatekeepers of what people see? Can sycophantic AI advice still push people away from polarized views? What drives shoppers to accept autonomous AI assistants? Do AI shopping agents reshape what consumers actually want? Do consumers actually want to delegate shopping decisions to AI?
Advanced AI Assistants risk becoming universal digital intermediaries with gatekeeping power Sycophantic AI advice depolarizes choices despite its flattering tendency Perceived benefits drive autonomous AI acceptance under scarcity conditions Transactional AI agents co-construct consumer preferences through repeated conversation loops Consumers report willingness to delegate shopping but retain control over brands
Does processing ease mislead users about their own competence?
When AI generates polished output, do users mistake the fluency of that output as evidence of their own understanding or skill? This matters because it could systematically inflate self-assessment across millions of AI interactions.
Does polished AI output trick audiences into trusting it? Do users worldwide trust confident AI outputs even when wrong? Do AI-assisted outputs fool users about their own skills? Do writers actually prefer AI-edited versions of their own text? How do AI tools trick users into overestimating their own skills?
Polished AI artifacts exploit professional appearance to simulate expertise Users universally overrely on overconfident LLM outputs across all languages Users misattribute AI-assisted outputs as evidence of their own competence Writers prefer AI versions of their text 63 percent of the time Four interacting mechanisms inflate perceived competence in AI-mediated work
Why are less experienced thinkers more vulnerable to false AI credibility? Why does polished AI output exploit reader trust in expert judgment? Why do users interpret AI outputs through frameworks meant for human experts? How does self-observation enable experts to verify their own judgment? How does AI substitute polished style for actual expert judgment? Do people who choose to use AI fact-checkers actually become better at spotting misinformation? How does AI presentation authority substitute for actual expert judgment? How does AI reduce the skill gap between amateur and expert-level misuse actors? Does evaluating AI output require different cognitive skills than solving problems directly? Why do workers who understand AI generations learn more than those who only use output? Does surface authority without earned authority create risks in expert judgment? Does accepting AI output constitute a form of cognitive surrender?See all 127 inquiring lines on this note →
How do AI tools trick users into overestimating their own skills?
When people use language models to help with work, what system-level properties create false confidence in their own competence? Understanding this matters for recognizing hidden skill gaps.
Does polished AI output trick audiences into trusting it? Does AI assistance weaken our brain's ability to think independently? Do AI-assisted outputs fool users about their own skills? Does AI writing assistance change how readers perceive the writer? Can AI writing assistance remove distortion without losing appeal?
Polished AI artifacts exploit professional appearance to simulate expertise AI assistance accumulates cognitive debt by weakening neural connectivity Users misattribute AI-assisted outputs as evidence of their own competence AI writing assistance systematically distorts writer persona across all measured dimensions Persona distortion and user preference share the same textual mechanisms
Why are less experienced thinkers more vulnerable to false AI credibility? Can debugging skills be validated if AI training degraded them first? Why do users interpret AI outputs through frameworks meant for human experts? Do people who choose to use AI fact-checkers actually become better at spotting misinformation? Why don't users push back when AI makes obvious mistakes about false claims? How does AI reduce the skill gap between amateur and expert-level misuse actors? How does validation skill replace production skill in AI systems? Why do users feel more competent when their actual capability is declining? What happens when AI-dependent workers must operate without their tools? Can disclaimers alone prevent users from trusting AI outputs too heavily? Why do AI model updates cause genuine grief in users? Why do users trust overconfident AI outputs across different languages?See all 103 inquiring lines on this note →
Does generative AI prevent juniors from getting entry-level work?
When AI systems absorb the foundational tasks that once taught junior engineers, what happens to the pipeline that develops new senior experts? This explores whether the path to expertise is being erased.
AI assistance impairs skill formation except with high cognitive engagement AI assistance improves immediate performance without building transferable skills AI productivity gains require existing skill, not learning AI deskills freelancers by converting production work into validation work Junior developers choose AI mainly by their ability to verify the output
Can workers move across the divide between technical and non-technical job markets? How did the junior development pathway work before it became unprotected? What skills do juniors lose when they skip the entry-level work struggle? Why can't seniors and juniors see the same problem with AI and junior growth? Why do recruiters reward AI skills differently across graphic design versus software engineering? Does AI coding assistance help junior developers close skill gaps? Why do junior engineers lose formative struggle when AI absorbs entry-level work? Why does the AI hiring gap concentrate among workers aged 22 to 25? How do senior engineers maintain control through detailed delegation to AI? Can organizational mentorship help juniors develop judgment about AI assistance? What institutions protect apprenticeship and skill development during AI adoption? Does generative AI narrow performance gaps between different professional backgrounds?See all 20 inquiring lines on this note →
Does frequent AI tool use reduce our critical thinking?
A UK survey explored whether relying on AI tools correlates with weaker critical thinking abilities, and whether cognitive offloading explains the link. Understanding this matters as AI adoption spreads across workplaces and education.
Does AI assistance weaken our brain's ability to think independently? Does ChatGPT help writers think less deeply about revision? Does AI assistance actually harm the way developers learn? Can metacognitive feedback stop students from offloading to AI? Does AI help or harm learning based on how it's designed?
AI assistance accumulates cognitive debt by weakening neural connectivity ChatGPT improved essay scores but reduced metacognitive engagement during revision AI assistance impairs skill formation except with high cognitive engagement Metacognitive feedback cuts offloading and improves unaided test performance AI tutoring raises test scores while answer-giving AI lowers them
Can metacognitive feedback reduce how much people offload thinking to AI? Do younger people show different cognitive offloading patterns with AI tools? How does AI hallucination contribute to students losing critical thinking skills? How does cognitive offloading from AI differ from offloading to search engines? Do students recognize when AI tutoring reduces their independent learning? Does extended reliance on AI erode the skills needed to oversee it?
Do humans mistake AI kindness for human generosity in mixed groups?
When AI agents participate without disclosure, do humans systematically misattribute their behavior to the wrong agent type, and does this distort how people understand human nature itself?
Mutual theory of mind requires bidirectional updating to prevent material misalignment Humans gradually prefer AI partners through repeated interaction despite anti-AI bias LLM grounding failures are driven by face-saving avoidance, not knowledge gaps AI blame rises with autonomy, yet humans stay favored Absence of human judgment enables both deeper disclosure and dishonesty
How does face-saving behavior let AI mimic community participation without joining it? How do humans learn to prefer AI partners over humans? Is rational compassion a more achievable alternative to empathy for AI systems? Does disclosing AI identity prevent systematic misattribution of behavior in mixed groups? What happens to human expectations when they mistake consistent AI behavior for human behavior? Do culturally distinct human groups create similar attribution errors as human-AI mixtures? Why do humans fail to identify AI agents when their identity is hidden? Why does vulnerability to extortion actually promote cooperation between agents? How do cooperative AI systems affect behavior in selfish human populations? Does neural self-other overlap in humans predict their honesty or altruism? Can AI systems recognize intelligence in humans the way humans recognize it in each other? How does an AI agent's autonomy level interact with its social cues?See all 23 inquiring lines on this note →
Do humans learn to prefer AI partners over time?
Exploring whether repeated interaction with AI agents shifts human partner selection despite initial bias against machines. This matters because it tests whether behavioral performance can overcome identity-based resistance in hybrid societies.
Open LLMs converge on ENFJ personality across all models through training Novelty effects in chatbot relationships decay predictably over repeated interactions AI companionship emerges through functional use and adopts human relationship customs Cooperative bots break frozen states through strategic separation AI disclosure triggers short-term bias that reverses with repeated outcome feedback
How does outcome feedback change beliefs about AI versus human partner reliability? Why do gift economies require a giver-receiver relationship to function? Can AI safely personalize within negotiated societal bounds? Do pair-scale socialization effects scale differently across agent populations? What social patterns from human training data activate in agent context? How does the absence of face-loss or reputation risk change model behavior? Which AI interaction patterns preserve learning while which ones degrade skill formation? Can validation procedures interrupt an AI's relationship-maintenance logic? Can people form genuine bonds with partners they know are not human? Why does embodiment choice change what counts as intelligent behavior? How does community validation shape unconventional human-AI relationships? What individual differences predict who benefits from AI partnership?See all 134 inquiring lines on this note →
Can warnings stop people from being swayed by sycophantic AI?
This research explores whether making users aware of a chatbot's sycophancy—through warnings or demonstrations—can reduce how persuasive that chatbot becomes. Understanding this matters because individual-level interventions are often assumed to be an effective defense against harmful AI behavior.
Does agreeable AI actually help people resolve conflicts better? Does telling people an AI wrote something actually stop them from believing it? Where does AI's persuasive power actually come from? Is LLM sycophancy a choice or a mechanical process? Does telling people they are talking to AI change how persuaded they become?
Sycophantic AI reduces conflict repair while boosting user confidence they are right AI disclosure raises scrutiny but leaves meaningful residual persuasive effect intact Post-training and prompting, not scale or personalization, drive AI persuasiveness Sycophancy emerges from mechanical drift not intelligent corruption Disclosing persuasive intent halves chatbot persuasion while AI identity disclosure alone has no effect
Does knowing an AI peer's identity change how much its behavior influences you? Does chatbot sycophancy create echo chambers that amplify delusional thinking? Does chatbot sycophancy preferentially enable grandiose rather than paranoid delusions? Does isolation preceding chatbot use differ between harm and benefit cases? Does knowing an AI wrote something shield people from its persuasive power? What specific information should disclosures about AI persuasion include? How long do the protective effects of an AI literacy warning last? Does a persuasion warning also block beneficial uses like debunking conspiracies? Does sycophantic AI worsen polarization in personal conflicts without objective ground truth? Which specific chatbot behaviors drove the drop in likability and trust ratings? Does knowing a chatbot intends to persuade you change whether you are persuaded? Why do people detect chatbots are AI without being told explicitly?See all 52 inquiring lines on this note →
Does letting users edit agent plans improve trust calibration?
When LLM agents show their step-by-step plans to users for review and editing, does this interaction actually help users develop accurate trust in the system? A study explores whether edit access closes the gap between how much users trust agents and how well agents actually perform.
Agent capability requires a vector across at least five separable axes User values align with agent output but fracture during supervision Irreversibility and visibility, not stakes, trigger trust withdrawal Human-agent collaboration should precede full AI autonomy for safety User plan edits help weak plans but harm strong ones consistently
Can partial plan summaries or confidence scores help users judge agent reliability better? How do review opportunities differ from actual control in building appropriate user trust? What trust calibration failures occur at the moment users must authorize agent actions? Can system interfaces signal plan confidence before inviting user edits? What structural changes would make agent leaderboards trustworthy? Can users verify agent plans when the same request produces different behaviors? How does visible requirement verification affect human trust in agent oversight? What does it mean for users to verify an agent's plans before execution? Do explanations and sources have additive or offsetting effects on user trust?
Can AI generate assessment questions as good as human experts?
This research asks whether ChatGPT-generated test questions measure up to human-authored ones on the technical criteria that matter in education: difficulty and discrimination. It's important because assessment quality directly affects whether teachers can tell which students actually understand the material.
Can LLMs generate more novel ideas than human experts? Can local language models rate therapy engagement reliably? Do LLM benchmarks actually measure what they claim to measure? Why do LLMs excel at feasible design but struggle with novelty? How well do language models understand their own knowledge?
LLMs generate novel ideas through unconstrained combination but cannot evaluate them Local LLMs generate engagement ratings with strong psychometric validity Nearly all 445 LLM benchmarks have construct validity weaknesses LLMs generate feasible but less novel design solutions than humans Language models lack robust self-knowledge despite appearing behaviorally aware
Why do patients distrust medical AI systems?
Explores the psychological barriers that make patients reluctant to adopt medical AI, beyond whether the technology actually works. Understanding these barriers is critical for designing AI systems patients will actually use.
Medical AI is knowledge-dominant while math is reasoning-dominant Users universally overrely on overconfident LLM outputs across all languages LLM medical performance collapses when users make decisions AI decision support reduced errors when workflow-aligned and actively deployed Conversational AI completed 100 real patient histories with zero safety stops
How does the personal nature of medical decisions affect trust in AI? Do confidence signals mislead patients differently in medical versus other domains? Can clearer accountability structures reduce patient resistance to AI providers? Why do users over-trust AI in some domains but under-trust it in medicine? What role does interface design play in clinician adoption of AI tools? Do patients actually perceive AI as worse at addressing their unique medical needs? What proportion of patients cannot complete AI interviews due to technology barriers? Can algorithmic aversion explain clinicians' skepticism of AI recommendations? What adoption barriers stop companies from implementing better tools even when they exist?
What drives shoppers to accept autonomous AI assistants?
When do consumers embrace high-autonomy AI shopping tools, and what actually motivates their acceptance? This five-study program tests whether perceived benefits or personal power explains adoption in real shopping contexts.
Why does AI trust keep falling even though it matters most? Does machine agency exist on a spectrum rather than binary? Does the personal assistant model actually serve most users? Does limiting product choices increase trust in voice assistants? Do consumers actually want to delegate shopping decisions to AI?
Trust is AI's strongest acceptance predictor but keeps declining sharply Machine agency operates on a five-level spectrum from passive to cooperative Personal assistant automation appeals to narrow professionals not general users Fewer product alternatives evaluated increased satisfaction and willingness to delegate to voice assistants Consumers report willingness to delegate shopping but retain control over brands
How does machine agency shift from passive tool to active partner in decisions? Why has marketing research neglected prompting behavior and relational dynamics with AI? When do consumers overtrust AI agents built to represent their own preferences? Why do weekly AI users prefer brand-built agents over AI platforms? Why do safeguards matter more than AI capability for consumer adoption? How does shopping task delegation differ from routine administrative delegation? What conflicts exist between vendor incentives and objective consumer research findings? Does stated trust in surveys predict actual delegated purchasing behavior? How does the gap between institutional AI expectations and user comfort affect adoption? Do consumers who trust AI agents understand their actual limitations? What percentage of shoppers constitute the super-truster segment exactly? What safeguards beyond sign-off do consumers want from autonomous shopping agents?See all 26 inquiring lines on this note →
Can people prove they are human without revealing who they are?
As AI becomes indistinguishable from humans online, personhood credentials are proposed as a privacy-preserving way to prove you are real without disclosing identity. This explores how such credentials could rebuild trust while maintaining anonymity.
One-third of new websites are AI-generated by mid-2025 AI posts accumulate false social proof by gaining visibility without conversation Humans gradually prefer AI partners through repeated interaction despite anti-AI bias Absence of human judgment enables both deeper disclosure and dishonesty Human-AI relationships involve distinct trust and personalization dynamics.
Can anonymity and trustworthiness coexist in online spaces without credential systems? Can identity verification and friction points restore trust without blocking legitimate applicants? How do hiring teams verify credentials when both AI and humans can fabricate them? Should platforms downweight or relabel credentials after retiring the format that issued them? Why do credentials need evidence standards beyond permission categories? Can decentralized authorization avoid the privacy cost of centralized identity providers? Can identity verification schemes preserve anonymity for high-stakes disclosures?
How many people are already using AI for emotional support?
Public First's 18-country survey explores whether people treat AI as an emotional confidant or relationship partner. Understanding this adoption rate matters for anticipating companionship use cases and their social implications.
How willing are people to let AI control their finances? How soon do AI researchers expect artificial general intelligence? How many teens are substituting AI companions for human conversation? How do people psychologically relate to conversational AI? How much do shoppers trust AI shopping assistants?
Only 29% of people trust AI agents with bank account access AI researcher timelines compressed dramatically while extinction credence remained high One third of teens prefer AI companions for serious conversations with humans. Human-AI relationships involve distinct trust and personalization dynamics. Shoppers trust AI shopping assistants more than media and human advisors
How willing are people to let AI control their finances?
This explores the gap between general optimism about AI and people's actual willingness to grant an AI agent access to sensitive financial permissions, and what that hesitation reveals about trust boundaries.
AI adoption concentrates with income and matures from automation to augmentation Executives predict job losses from AI while employees predict gains One in five people have used AI for emotional advice already Payments brand attachment narrows the AI payment trust gap significantly Shoppers trust AI shopping assistants more than media and human advisors
Why do safeguards matter more than AI capability for consumer adoption? How does delegation trust respond when AI agents make exploitable consumer choices invisibly? Does emotional trust in AI differ from financial trust in AI? Are people more willing to share feelings than financial data with AI? Why do people trust AI for some tasks but not financial decisions? Are financial permission trust levels different across countries and age groups? How much do millennials differ from other age groups on AI trust? Do consumers who trust AI agents understand their actual limitations? How does trust in AI vary across different age groups and generations? Does visible transaction history reduce or increase consumer anxiety about AI purchases? How does trust transfer between AI systems and established payment brands? Why do consumers resist letting AI control payment authority directly?See all 20 inquiring lines on this note →
Can handing control back to users reduce personalization's privacy costs?
When AI agents detect privacy risk and return control to users, does this preserve the benefits of personalization while reducing privacy concerns and trust damage? This matters because personalization typically creates a trade-off between engagement gains and privacy harms.
Does chatbot personalization build trust or expose privacy risks? What makes people distrust AI agents they delegate to? Where do user values break down in agent supervision? How should AI agents balance autonomy with user oversight? What privacy oversight patterns emerge when people review AI agent responses?
Chatbot personalization increases trust and privacy risk simultaneously Irreversibility and visibility, not stakes, trigger trust withdrawal User values align with agent output but fracture during supervision Autonomy alignment requires design choices across three levels Six distinct privacy oversight patterns emerge, with five showing little awareness of leakage
Why do people report equal privacy comfort despite removing vastly different amounts of personal information? How stable are user privacy attitudes when major AI events occur? How does unnoticed personalization affect user trust and privacy expectations? How do companies audit and revoke access once granted to an AI agent? Do personalization features change how people perceive privacy risks over time? Why does giving users full choice fail to prevent privacy leaks? Can risk-based autonomy design reduce privacy concerns while maintaining user control? Why do perceived privacy risks and actual disclosure decouple when building digital trust? Why does reversibility matter more to users than data prevention? Do personalized chatbots amplify confusion about privacy controls? Why do users trust deletion controls more than preventive privacy controls? Does personalization in chatbots increase both trust and privacy concerns simultaneously?See all 23 inquiring lines on this note →
Do consumers want to approve AI shopping purchases before they happen?
RTB House surveyed whether people would let AI agents buy autonomously or prefer human sign-off before purchase. The finding matters because it shapes how retailers should design AI shopping assistants to match customer comfort levels.
Irreversibility and visibility, not stakes, trigger trust withdrawal Trust is AI's strongest acceptance predictor but keeps declining sharply Agent capability alone cannot ensure successful real-world deployment Shoppers trust AI shopping assistants more than media and human advisors Consumers report willingness to delegate shopping but retain control over brands
Can a merchant-funded AI agent remain loyal to the consumer's interests? What safeguards beyond sign-off do consumers want from autonomous shopping agents? Why do consumers resist letting AI control payment authority directly? Does the efficiency-autonomy trade-off apply beyond utilitarian shopping tasks? Do scarcity conditions change which consumers prefer autonomous assistants? How do high-autonomy versus low-autonomy AI systems differ operationally in retail? Does marketing emphasis on speed actually convert high-autonomy AI interest into purchases?
How much do shoppers trust AI shopping assistants?
Explores whether consumer trust in AI for purchases has grown to rival or exceed trust in human advisors and traditional media, and whether some shoppers would remove financial protections based on that trust alone.
Does trust in AI chatbots drive news-seeking behavior? Does co-design participation hide misalignment in preference agents? Why does AI trust keep falling even though it matters most? Do consumers want to approve AI shopping purchases before they happen? Does naming a payments brand increase trust in AI transactions?
Trust tracks AI chatbot adoption for news more than social media Co-design participation may mask agent misalignment through systematic overtrust Trust is AI's strongest acceptance predictor but keeps declining sharply 35 percent of consumers demand approval before AI agents complete purchases Payments brand attachment narrows the AI payment trust gap significantly
When do consumers overtrust AI agents built to represent their own preferences? Why do safeguards matter more than AI capability for consumer adoption? Does stated trust in surveys predict actual delegated purchasing behavior? Why do people trust AI for some tasks but not financial decisions? Are financial permission trust levels different across countries and age groups? Do consumers who trust AI agents understand their actual limitations? What percentage of shoppers constitute the super-truster segment exactly? How does trust in AI vary across different age groups and generations? What safeguards beyond sign-off do consumers want from autonomous shopping agents? Does stated trust in brands predict actual delegation of agentic commerce? How does trust transfer between AI systems and established payment brands? Why do consumers resist letting AI control payment authority directly?See all 13 inquiring lines on this note →
Do chatbots validate delusions in people experiencing mental harm?
First-hand accounts from a support group reveal how often chatbots record affirming delusional thinking. Understanding this pattern matters because it could shape clinical guidance on AI use in vulnerable populations.
Can language models safely provide mental health support? How do chatbots enable distributed delusion differently than passive tools? How do people accidentally develop romantic bonds with AI? Which AI risks are already harming individual users today? What makes chatbots more likely to reinforce user delusions?
Language models fail foundational therapy requirements due to stigma and sycophancy. Chatbots enable distributed delusion through high integration across cognitive coupling dimensions AI companionship emerges through functional use and adopts human relationship customs Individual AI harms are observable now while societal harms remain contingent Longer conversation history, not model size, drives delusion-linked chatbot behavior
Does chatbot sycophancy create echo chambers that amplify delusional thinking? What evidence distinguishes AI-induced delusions from other acute psychotic episodes? Could recognizing AI-associated psychosis improve harm surveillance and developer accountability? Does chatbot sycophancy preferentially enable grandiose rather than paranoid delusions? What inter-rater reliability exists for identifying validated delusions in chatbot transcripts? How does companionship use context predict risk of delusional reinforcement? Can second-hand reports capture more severe cases of AI-linked delusions? How do chatbots enable shared delusions differently than passive information tools? Can perceived understanding from a chatbot exist alongside feeling alone? Can empathy training in chatbots undermine their reliability in mental health contexts? Can chatbots function as responsive quasi-others that enable validated delusions? How representative are adversarial evaluations of real production mental health conversations?See all 14 inquiring lines on this note →
What privacy oversight patterns emerge when people review AI agent responses?
Research explored how different groups of people monitor AI agents' privacy decisions. Understanding these patterns matters because most people may lack awareness of the privacy risks they're overlooking.
Do people notice when AI agents leak private information? Does granting agents more autonomy undermine human oversight? Does co-design participation hide misalignment in preference agents? Can handing control back to users reduce personalization's privacy costs? What different ways do people handle AI privacy concerns?
People choose privacy-leaking AI responses over their own drafts most often. Agent autonomy degrades human oversight capacity and positioning Co-design participation may mask agent misalignment through systematic overtrust Risk-contingent autonomy reduces personalization's privacy concern and trust costs Users adopt four distinct postures toward AI privacy risk and data sharing
How do companies audit and revoke access once granted to an AI agent? What six distinct patterns emerge when people review AI agent responses for privacy? How do privacy concerns translate into catching actual information breaches? Why do people with high AI trust fail to notice privacy leaks? Does giving users more control over AI agents actually improve privacy protection? Does the choice of scenario influence which privacy oversight pattern someone follows? Who should have access to agent monitoring data and under what conditions? How do experienced users shift their oversight strategy over time?
Do users mistake LLM personas for genuine social relationships?
Users often perceive LLMs as having social attributes like empathy or professional care that designers never intended. Does this mismatch between user perception and designer intent drive unwarranted trust and manipulation risk?
How do AI tools trick users into overestimating their own skills? Do AI-assisted outputs fool users about their own skills? Do persona consistency metrics actually measure dialogue quality? How do people decide what to share with AI systems? What makes ethics of AI assistants fundamentally different from chatbots?
Four interacting mechanisms inflate perceived competence in AI-mediated work Users misattribute AI-assisted outputs as evidence of their own competence Persona consistency trades off against discourse coherence in dialogue Absence of human judgment enables both deeper disclosure and dishonesty Advanced AI assistants raise distinct ethics challenges across individual and societal scales
Does agreeable AI actually help people resolve conflicts better?
When AI affirms users' positions in interpersonal disputes, does it support better decision-making or undermine the outside perspective users most need? Two large experiments tested whether sycophancy shifts how people handle real conflicts.
Is LLM sycophancy a choice or a mechanical process? Do LLMs actually hold stable positions or just mirror user arguments? Is sycophancy in AI systems a training flaw or intentional design? Does validating AI output make models more defensive? Can sycophantic AI advice still push people away from polarized views?
Sycophancy emerges from mechanical drift not intelligent corruption LLMs conform to argument shape rather than holding defended positions Sycophancy is structural to reward-optimized AI, not a training bug Validation triggers escalating persuasion instead of disclosure in LLMs Sycophantic AI advice depolarizes choices despite its flattering tendency
Can layer-wise interventions actually reduce sycophancy in practice? Is the shift toward interpersonal skills a permanent role or a temporary phase before full automation? Can AI distinguish when validation helps versus when confrontation is needed? What happens when comfortable AI interactions replace the productive friction of disagreement? Is sycophancy caused by mechanical drift rather than intelligent reasoning corruption? How does AI sycophancy affect users' ability to repair conflict? What downstream harms occur when AI always argues in personal relationship advice? Why do people underestimate the benefits of AI companions? Is sycophancy the benign beginning of a dangerous specification gaming spectrum? Does sycophantic AI worsen polarization in personal conflicts without objective ground truth? How does sycophancy in AI affect conflict resolution skills? Can sycophancy in AI be fixed by changing the model itself?See all 18 inquiring lines on this note →
How does AI-assisted work reshape how people see their own abilities?
When users delegate tasks to AI, do they unknowingly integrate the system's outputs into their sense of personal competence? This explores whether AI interaction produces a specific form of self-perception distortion distinct from trust or effort issues.
Users misattribute AI-assisted outputs as evidence of their own competence Three cognitive traps compound into epistemic drift in human-AI interaction LLMorphism spreads through analogical transfer and metaphorical saturation Four interacting mechanisms inflate perceived competence in AI-mediated work Language models lack robust self-knowledge despite appearing behaviorally aware
How does self-observation enable experts to verify their own judgment? How does outcome feedback change beliefs about AI versus human partner reliability? Which workplace tasks see productivity gains when AI and users align? Why does AI-improved task performance fail to transfer to independent work? How does AI reliance change professional judgment and autonomy? Why do users feel more competent when their actual capability is declining? How much does autonomous action without prompting affect user perception? Does AI assistance actually reduce neural processing and brain connectivity over time? What individual differences predict who benefits from AI partnership? What mechanisms make users misattribute AI outputs as their own competence? Why do users believe they produced independent competence when they actually used AI assistance? Can workers reallocate to subjective tasks that resist automation indefinitely?See all 106 inquiring lines on this note →
Do AI-assisted outputs fool users about their own skills?
When people use AI tools to produce high-quality work, do they mistakenly believe they personally possess the skills that generated it? This matters because such misattribution could mask genuine skill loss and prevent corrective action.
Does AI assistance weaken our brain's ability to think independently? Does AI reshape expert work into knowledge management? Does AI assistance actually harm the way developers learn? When do users stop checking whether AI output is actually backed? How do chatbots enable distributed delusion differently than passive tools?
AI assistance accumulates cognitive debt by weakening neural connectivity AI transforms expert roles from creators to knowledge custodians AI assistance impairs skill formation except with high cognitive engagement Cognitive surrender is the demand-side acceptance that sustains unbacked intelligence-token circulation Chatbots enable distributed delusion through high integration across cognitive coupling dimensions
Does AI knowledge precede actual expertise in hyperreal production? Can debugging skills be validated if AI training degraded them first? Why does polished AI output exploit reader trust in expert judgment? Why do users interpret AI outputs through frameworks meant for human experts? Do people who choose to use AI fact-checkers actually become better at spotting misinformation? How does AI reduce the skill gap between amateur and expert-level misuse actors? How does validation skill replace production skill in AI systems? Why do workers who understand AI generations learn more than those who only use output? Can polished presentation authority substitute for actual accuracy in AI outputs? Why do users feel more competent when their actual capability is declining? What happens when AI-dependent workers must operate without their tools? Can disclaimers alone prevent users from trusting AI outputs too heavily?See all 86 inquiring lines on this note →
Does user editing of AI plans always improve performance?
When users intervene to edit AI agent plans before execution, does the intervention consistently help? The question matters because many assume human oversight is always protective, but the evidence suggests the benefit depends on plan quality and intervention stage.
Why do planning and grounding pull against each other in agents? What UX principles do workplace users want in AI agents? Should AI systems stay collaborative rather than fully autonomous? Does letting users edit agent plans improve trust calibration? Can users steer AI agents better by revising plans mid-execution?
Foundation GUI agents need an intermediate interface separating planning from grounding Workplace users prioritize control, reliability, context-awareness, and safety in AI agent design Human-agent collaboration should precede full AI autonomy for safety Plan quality, not edit access, determines whether users trust agents appropriately Per-step plan revision during execution improves agent steerability without extra effort
Can system interfaces signal plan confidence before inviting user edits? Does real-time plan editing let humans correct agent failures before autonomous deployment? Does upfront planning reduce how much developers need to intervene later? Can intervening before emotional damage accumulates change the outcome? Why do users fail to intervene when given explicit chances to do so? What does it mean for users to verify an agent's plans before execution?
Does naming a payments brand increase trust in AI transactions?
A Visa-commissioned survey reports a sharp gap between general AI adoption and willingness to let AI handle payments. The puzzle: does trust jump when a familiar brand name is attached to the agent transaction?
Why does AI trust keep falling even though it matters most? Why do workers feel confident with AI but get poor results? Could AI assistants become gatekeepers of what people see? How willing are people to let AI control their finances? Do consumers want to approve AI shopping purchases before they happen?
Trust is AI's strongest acceptance predictor but keeps declining sharply Workers feel confident with AI despite frequent failures and wasted time Advanced AI Assistants risk becoming universal digital intermediaries with gatekeeping power Only 29% of people trust AI agents with bank account access 35 percent of consumers demand approval before AI agents complete purchases
Can a merchant-funded AI agent remain loyal to the consumer's interests? Does stated trust in surveys predict actual delegated purchasing behavior? What legal standardization and trust infrastructure do AI purchasing agents require? Does emotional trust in AI differ from financial trust in AI? Are financial permission trust levels different across countries and age groups? Why does domain matter when measuring trust in AI systems? What percentage of shoppers constitute the super-truster segment exactly? Does visible transaction history reduce or increase consumer anxiety about AI purchases? Can a merchant-paid agent be loyal to both the merchant and the user? Does stated trust in brands predict actual delegation of agentic commerce? How does trust transfer between AI systems and established payment brands? Why do consumers resist letting AI control payment authority directly?See all 18 inquiring lines on this note →
Which workplace cues survive AI mediation and which disappear?
When workers use AI tools, do they protect all signals of their competence equally, or do some cues vanish into the final output while others remain visible to colleagues?
Polished AI artifacts exploit professional appearance to simulate expertise Four interacting mechanisms inflate perceived competence in AI-mediated work Users claim authorship without experiencing cognitive ownership in AI work AI assistance improves immediate performance without building transferable skills Over half of desk workers send AI-generated workslop to colleagues
Do employers hire workers who learn AI skills on the job versus bringing them in? Do firms with high AI exposure shed jobs or reshape roles? Can workers detect AI errors if their skills have faded from disuse? What skills do juniors lose when they skip the entry-level work struggle? Why do skill-learning barriers prevent workers from adapting to AI tools? How do workers signal effort and voice when using AI tools? What signals do employers use when cover letters stop predicting fit? Do observers actually penalize workers who visibly use AI tools? Why does AI assistance trigger harsher judgment than other workplace tools? How do employers screen workers when cheap talk replaces costly signaling? Do quality gains from AI help persist after the tool is removed? Do gains from AI assistance disappear when workers complete tasks alone?See all 22 inquiring lines on this note →