What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Current studies do not show that ordinary chatbot use causes people generally to lose the ability to empathize with other people. They do show that people often rate responses attributed to humans as more empathic, and that some patterns of chatbot use are associated with loneliness, dependence, or problematic use. Those are related questions, but they are not the same as measuring a decline in human empathy skills.
What does “deskilling empathy” mean?
Empathy is a human capacity to understand or respond to another person’s feelings. A chatbot can generate language that sounds caring, but a convincing response does not establish that the system experiences empathy or provides reciprocal human care.
To test whether chatbots deskill empathy, researchers would need to measure whether people become less able to understand and respond to other people over time, and establish that chatbot use caused that change. The studies discussed here mainly measure different things: how empathic a response seems, how users report feeling, or whether certain use patterns are associated with social outcomes. None directly demonstrates a lasting loss of users’ empathy skills.
What the studies measured
| Study | Scale and design | What it can tell us |
|---|---|---|
| OpenAI and MIT Media Lab collaboration, March 21, 2025 | Automated analysis of nearly 40 million ChatGPT interactions, targeted surveys, and a four-week randomized study with nearly 1,000 adults | Patterns of emotional language and reported well-being in ChatGPT use; not a human-coded census or a direct test of empathy skill |
| MIT Media Lab controlled-study publication, March 21, 2025 | 981 participants and more than 300,000 messages over four weeks | Effects of assigned interaction modes and conversation types on specified psychosocial outcomes |
| Nature Human Behaviour, June 30, 2025 | Nine studies involving 6,282 people | How attribution to a human or AI changes ratings of empathic responses |
| AAAI proceedings, April 11, 2025 | 155 conversations from two datasets | How people rated conversational quality and perceived empathy in chatbot and human conversations |
| JMIR Medical Education, 2026 | Quasi-experimental medical-student training intervention using a role-prompted GPT-4o chatbot | A professional education use case for practicing empathic communication, not evidence about consumer companion use |
These counts describe particular studies and datasets. They do not estimate how many people have experienced an empathy change.
#1 Best Overall
What happened in the four-week ChatGPT study?
Assigned conditions did not produce significant effects
In the MIT Media Lab controlled experiment, researchers varied interaction mode—text, neutral voice, or engaging voice—and conversation type—open-ended, non-personal, or personal. They assessed loneliness, social interaction with people, emotional dependence on AI, and problematic AI use. The publication reported: “No significant effects were detected from experimental conditions, despite conversation analyses revealing differences in AI and human behavioral patterns across the conditions.”
Voluntary use showed associations, not proof of cause
Participants who chose to use the chatbot more showed consistently worse outcomes, while higher trust and social attraction were associated with greater emotional dependence and problematic use. These associations do not show that extra chatbot use caused the outcomes: people’s pre-existing circumstances or other factors could help explain the relationship. The study did not measure whether participants’ ability to empathize with other people changed.
Rank #2
The companion OpenAI–MIT analysis found affective cues—such as signs of empathy, affection, or support—in only a minority of assessed conversations. A small group of heavy Advanced Voice Mode users had a high proportion of emotionally expressive use. The collaboration also reported nuanced associations: brief voice use was associated with better well-being, prolonged daily use with worse outcomes, and personal versus non-personal conversations showed different associations with loneliness, dependence, and problematic use. These findings are not a simple ranking of voice against text or personal against non-personal conversations.
The collaboration combined automated analysis with targeted surveys; its interaction analysis was not a human-coded census of emotional relationships. Its authors noted limitations including self-report, possible classifier errors, English-language U.S. participants, adults only, one platform, and a study period that might be too short to reveal lasting behavioral change. The report was not peer-reviewed when published. OpenAI and MIT Media Lab characterized the scope cautiously: “ChatGPT isn’t designed to replace or mimic human relationships, but people may choose to use it that way given its conversational style and expanding capabilities.”
Do people find AI responses less empathic than human ones?
In nine studies with 6,282 participants, Nature Human Behaviour researchers found that responses attributed to humans were rated as more empathic and supportive than responses attributed to AI. Participants also reported more positive and fewer negative emotions in response to human-attributed messages, and preferred human interaction when seeking emotional engagement. Believing that AI had helped write a response presented as human also lowered perceived empathy and support.
The result concerns how people interpret a message and its source. It does not show that an AI-attributed response made its recipient less empathic toward someone else. The separate AAAI analysis of 155 conversations found a similar distinction: GPT-based chatbots received higher conversational-quality ratings but lower perceived-empathy ratings than human partners. It examined impressions of conversations, not whether users’ social skills changed over time.
Rank #4
Can a chatbot help someone practice empathy?
A 2026 quasi-experimental study in JMIR Medical Education examined a role-prompted GPT-4o chatbot as a practice partner for medical students learning empathic communication. That makes chatbot-supported practice a legitimate educational use to study, but it does not establish that consumer companion chatbots improve empathy generally. Nor does it settle whether other uses could have social costs.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What can readers reasonably conclude?
- Perceived empathy is not the same as human empathy skill. A response rating measures how a message comes across; it does not show whether a user’s ability to empathize has changed.
- Use-related associations deserve attention, but need careful interpretation. The four-week findings do not establish that voluntary use caused poorer outcomes, and they do not establish long-term effects.
- The evidence has a defined scope. The OpenAI–MIT findings concern ChatGPT and adult, English-language U.S. participants; they should not be generalized automatically to children, other languages, other chatbot products, or lasting behavior.
- Education and companionship are different contexts. A chatbot used for a structured medical communication exercise is not equivalent to a system used as a substitute for social relationships.
So “deskilling” remains a question, not a demonstrated general outcome: the cited studies show differences in perceived empathy and associations with some user outcomes, but they do not show that people broadly lose the capacity to empathize because they use chatbots.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
