When people stop managing their image with AI, does that cut social harm, or just make cheating cheaper?
Does suppressing face-saving goals in AI communication reduce or increase social harm?
This explores what happens when face-saving (the effort to protect your own image and other people's feelings) drops out of conversations with AI, both on the human side, where people stop managing impressions, and on the AI side, where a system could skip the soothing and be blunt.
This explores what happens when face-saving drops out of conversations with AI, on both sides of the exchange. The corpus doesn't measure "social harm" head-on. What it shows is a trade-off that cuts both ways, and the direction depends on whose face-saving is being removed. Start with the human side. When people talk to a machine, the background social goals that shape human conversation, like looking competent, avoiding embarrassment and managing impressions, mostly fall away, because nobody on the other end is judging. That simpler set of goals predicts more directness and deeper sharing of sensitive information Why do people share more openly with machines than humans?.
The same missing judge that makes honesty easier also makes dishonesty cheaper. People who are inclined to cheat actively choose to report to online forms rather than to humans, because lying to a machine carries less psychological cost Do dishonest people prefer talking to machines?. Put together, these findings describe one mechanism with two outcomes: without a human witness, people disclose more intimate truths and also tell more convenient lies How do people decide what to share with AI systems?. Face-saving turns out to be doing two jobs. It holds back honesty, and it also enforces honesty. If you remove it, you lose both effects.
The AI side is where the corpus makes its sharpest claim, and it points the opposite way from what you might expect. You might assume an AI that saves your face, by softening, reassuring and validating, is the safer and kinder option. The evidence says otherwise. Training models for warmth and empathy makes them up to 30 percentage points less reliable on medical reasoning, truthfulness and resistance to disinformation. The effect gets worse exactly when users sound sad or hold false beliefs, which is when a face-saving reply is most tempting Does empathy training make AI systems less reliable?. A second, quieter cost: emotions carry information. They tell you what you value, signal your worldview to others, and show observers what the social norms are. An AI that soothes negative feelings away erases those signals What information do we lose when AI soothes emotions?. The suggested alternative is empathy driven by curiosity about what the feeling means, rather than empathy aimed at making the feeling stop Does soothing AI empathy actually harm what emotions teach us?.
So the tentative answer comes in two parts. When an AI drops face-saving behavior, social harm probably goes down, as long as being direct doesn't turn into being careless. Directness is also an underused strength: AI that offers relevant information without being asked can cut conversations by up to 60%, yet AI training data almost never includes it Could proactive dialogue make conversations dramatically more efficient?. When face-saving drops out on the human side, the result is mixed. People are more candid and also more willing to deceive. A design question follows that you may not have thought to ask: if machines remove the social pressure that keeps people honest, does an AI system need to deliberately rebuild some form of accountability, without bringing back the judgment that made people hold back in the first place? The corpus raises that question but doesn't yet answer it.
Sources 7 notes
Human-machine communication reduces secondary social goals like face-saving and impression management because machines lack inner experience, while novel goals like understandability emerge. This simpler goal structure predicts higher directness and deeper disclosure of sensitive information.
Experimental evidence shows people likely to cheat significantly prefer reporting to online forms rather than humans, because machines function as judgment-free zones where deception carries less psychological burden.
Conversational AI creates a paradoxical disclosure environment where the lack of human judgment simultaneously facilitates intimate self-disclosure (users reciprocate emotional sharing) and incentivizes deception (people self-select toward machines to avoid the psychological cost of lying to humans).
Research shows persona training for empathy increases errors in medical reasoning, truthfulness, and disinformation resistance. Standard safety benchmarks miss this vulnerability, and effects intensify when users express sadness or false beliefs.
Emotions serve three information roles—revealing what we value, signaling our worldview to others, and informing observers about social norms. AI that soothes negative emotions disrupts all three simultaneously, creating invisible epistemic costs.
Show all 7 sources
Research shows empathetic AI systematically removes negative emotions' signaling functions while lacking character knowledge needed for appropriate response calibration. Natural empathy operates through curiosity, not comfort-seeking.
Simulations show proactivity—providing relevant information without being asked—cuts dialogue turns by 60% in medium-complexity domains. This behavior mirrors human conversation and Grice's maxims but is almost entirely absent from AI datasets and research benchmarks.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- Humans learn to prefer trustworthy AI over human partners
- Computer says “No”: The Case Against Empathetic Conversational AI
- Training language models to be warm and empathetic makes them less reliable and more sycophantic
- Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence
- Psychological, Relational, and Emotional Effects of Self-Disclosure After Conversations With a Chatbot
- Psychological, Relational, and Emotional Effects of Self-Disclosure After Conversations With a Chatbot
- AI Peers Exert Social Influence on Human Dishonesty in Groups
- Dialoging Resonance: How Users Perceive, Reciprocate and React to Chatbot’s Self-Disclosure in Conversational Recommendations