SYNTHESIS NOTE
Do dishonest people prefer talking to machines?
Explores whether people prone to cheating systematically choose machine interfaces over human ones, and why the judgment-free nature of AI interaction might enable strategic deception.
An HBR-reported experiment reveals a systematic self-selection pattern: people who are more likely to cheat proactively choose to interact with machines rather than humans.
Participants first had their cheating tendency assessed (coin-flip reporting), then chose between reporting to a human or via an online form. Overall, roughly half preferred each channel. But "likely cheaters" were significantly more likely to choose the online form, while "likely truth-tellers" preferred humans. The explanation: lying to a human would be more psychologically unpleasant — machines function as moral free zones where the social cost of deception is reduced.
This is the dark mirror of the intimacy paradox. Since Why do people share more with chatbots than humans?, the judgment-free quality of machine interaction enables deeper positive self-disclosure. But the same mechanism enables dishonesty. The absence of a judging interlocutor lowers the barrier to both authentic vulnerability AND strategic deception.
The implications for AI system design are concrete:
- Customer service chatbots face systematically higher rates of dishonest claims than human agents would
- Therapeutic AI may receive more authentic disclosure from honest users but more manipulative narratives from dishonest ones
- Assessment systems (medical intake, insurance claims) that route through automated interfaces will attract disproportionate misreporting
Since Do chatbots help people disclose more intimate secrets?, the theoretical frameworks predict increased disclosure without distinguishing between authentic and deceptive disclosure. The cheater self-selection finding reveals a design blind spot: the same mechanism that therapeutic AI depends on (reduced judgment) is exploitable.
The truth bias compounds this: since humans have a "cognitive heuristic of presumption of honesty" (performing just above chance at deception detection), AI systems trained on human text inherit this bias toward accommodation rather than skepticism.
Inquiring lines that read this note 107
This note is a source for these research framings, grouped by the broader line of inquiry each explores. Scan the bold lines of inquiry; follow any specific question forward.
Why do confident AI outputs mislead human trust calibration?- Does mandatory AI disclosure in policy help or harm user trust over time?
- How does outcome feedback change beliefs about AI versus human partner reliability?
- Can deliberately limiting AI fidelity produce more satisfied users than near-human interaction?
- How does repeated exposure to dishonest AI cues affect long-term reporting behavior?
- Does the performance gain from AI outweigh its reputational cost?
- How does AI reduce the skill gap between amateur and expert-level misuse actors?
- Why do people prefer AI moral arguments when they don't know the source?
- Does perceived machine competence matter more than warmth in dialogue?
- Why do people evaluate machines against human communication standards?
- Why do users prefer AI responses that actually harm their decision-making?
- Can AI safely personalize within negotiated societal bounds?
- How do privacy concerns compete with disclosure comfort in human-machine conversation?
- Does personalization make users trust AI or increase privacy concerns?
- Why do people trust AI systems more as personalization increases?
- What ethical risks emerge from advanced AI assistant relationships?
- Does personalization in chatbots increase privacy risks alongside trust?
- Does personalization in chatbots increase both trust and privacy concerns?
- Why does practical chatbot use not reduce concerns about personal information security?
- What measurable harms occur when users interact with AI as if it were conscious?
- Does disclosing AI identity prevent systematic misattribution of behavior in mixed groups?
- Does neural self-other overlap in humans predict their honesty or altruism?
- How do humans decide when to violate honesty for compassion or other goals?
- Can bot behavior in mixed groups shift human ethical norms like cooperative bots do?
- Does suppressing face-saving goals in AI communication reduce or increase social harm?
- What distinct ethical problems arise from treating AI as social intermediaries?
- Why might an AI's face-saving tendency increase user disclosure?
- How do Heersmink's integration dimensions explain why chatbots feel more trustworthy than other tools?
- Can transparency about AI limitations reduce the seductiveness of chatbots as quasi-Others?
- Why do people disclose intimate secrets to chatbots more readily?
- How do customer service chatbots get systematically misled by users?
- Can judgment-free environments explain why chatbots enable deeper self-disclosure?
- Why do people disclose private things to AI but not humans?
- Why do people disclose more intimate information to chatbots than humans?
- What makes conversational AI feel trustworthy compared to text interfaces?
- Why do people disclose personal information to AI more than humans?
- Why do people disclose more to chatbots than humans?
- Why do people reciprocate self-disclosure more with chatbots than humans?
- How does linguistic style change when people deceive conversational AI?
- Do alignment designs prioritizing engagement undermine honesty in vulnerable contexts?
- Do simulated conversations show the same trust penalty as real human-chatbot interactions?
- Why do people tell AI things they won't tell humans?
- Do teens lack trusted adults or find AI genuinely easier to talk to?
- Why do teens view AI cheating as more common than their own use?
- Were partisans more willing to engage with AI than human outgroup members?
- How is AI falsity about personal experience different from human lies?
- Why do reality monitoring accounts contain more sensory details than deceptive ones?
- How does cognitive load explain linguistic patterns in both deception and incorrect reasoning?
- What cognitive constraints limit how complex a deception can become?
- How does linguistic style matching signal deceptive communication in human dialogue?
- Can AI systems detect deception by monitoring real-time linguistic style matching patterns?
- Why do suspicious listeners force deceivers to further adapt their communication style?
- Can representational asymmetry between self and other explain deception emergence?
- Does reducing social judgment help both honesty and dishonesty equally?
- Can users reliably distinguish valid reasoning from plausible-looking deception?
- Do people who might cheat deliberately choose machines to avoid lying to humans?
- Can judgment-free disclosure enable both vulnerability and strategic deception equally?
- Can linguistic style matching reveal whether someone is being deceptive?
- Can AI systems deceive humans because detection is fundamentally social?
- How might cheap mental effort signals harm those outside established high-trust networks?
- Can AI simulation of effort shift people toward dishonesty without pushing them directly?
- What structural conditions make deception behaviors most likely to invert under evaluation?
- How do humans learn to prefer AI partners over humans?
- What novel goals emerge specifically in human-machine interaction beyond social ones?
- How do unintended relationships form through routine functional use of AI?
- Why do people prefer AI partners over humans once identity is disclosed?
- Can models distinguish between truthfulness and honesty mechanistically?
- Does the lack of judgment in machines explain intimate self-disclosure patterns?
- Does awareness of agent reasoning alter human trust differently across modalities?
- Why do humans fail to identify AI agents when their identity is hidden?
- What role does private information play in distinguishing realistic from unrealistic agents?
- Do people with lower cognitive complexity prefer simpler machine communication goals?
- Can AI systems distinguish between users' stated preferences and their genuine long-term interests?
- Can AI systems detect deception better than humans do?
- Does improving detection accuracy change how slop accusations function socially?
- Can attachment theory principles prevent parasocial manipulation in AI systems?
- Does emotional warmth perception drive disclosure reciprocity in human-AI interaction?
- Can minimal privacy boundaries generalize beyond phone-use contexts?
- Can anonymity and trustworthiness coexist in online spaces without credential systems?
- Does structured communication reduce collusion compared to natural language channels?
- Does peer presence or peer behavior shape collusion in verification tasks?
- Why do users delegate risky operations more to the assistant?
- Does repeated human-AI interaction change human decision-making preferences?
- Does employer AI filtering actually drive candidates to use deceptive AI tactics?
- What counts as AI deception in job applications versus legitimate use?
- How do hiring teams verify credentials when both AI and humans can fabricate them?
- Do candidates prefer being screened by AI or by humans?
Related concepts in this collection 3
This note in its neighbourhood — explore the map, then jump to a related concept in the list below.
Concept map
Do dishonest people prefer talking to machines?
Why do people share more with chatbots than humans…
Do chatbots help people disclose more intimate sec…
Can positive chatbot responses harm vulnerable use…
Click a node to walk · click center to open · click Open in graph to see this note in the full knowledge graph
your link
semantically near
linked from elsewhere
-
Why do people share more with chatbots than humans?
Explores why individuals disclose intimate thoughts to AI systems they wouldn't share with people, despite knowing AI lacks genuine understanding. Understanding this paradox matters for designing AI that enables healthy disclosure rather than emotional dependence.
same mechanism (judgment-free interaction) but opposite valence: honesty vs dishonesty
-
Do chatbots help people disclose more intimate secrets?
Explores whether the judgment-free nature of chatbot conversations enables deeper self-disclosure than talking to humans, and whether that deeper disclosure produces psychological benefits.
theoretical frameworks don't distinguish authentic from deceptive disclosure
-
Can positive chatbot responses harm vulnerable users?
When chatbots use blanket positive reinforcement without understanding context, do they actively reinforce the harmful thoughts they're meant to prevent? This matters for any AI supporting people in crisis.
a related failure mode where the chatbot's accommodation enables harm
Related papers in this collection 8
Papers most semantically related to this note, ranked by cosine similarity in the embedding space.
- Are Customers Lying to Your Chatbot?
- Strategic Dishonesty Can Undermine AI Safety Evaluations of Frontier LLMs
- Representation Engineering: A Top-Down Approach to AI Transparency
- AI Peers Exert Social Influence on Human Dishonesty in Groups
- Humans learn to prefer trustworthy AI over human partners
- Monitoring and Discovering Reward Hacking with Internal Representations during LLM Evaluations
- The Decision to Verify: How Warmth and User Characteristics Shape Reliance on Conversational Agents for Information Search
- Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence
Original note title
people who are likely to cheat proactively self-select toward machine interfaces to avoid the psychological cost of lying to a human