68/100

Moderate evidence

ProblemPublished recordLifestyleProblem Space

People offload empathy when they think AI can give it instead

Truvace record
TRV-2026-2502raw data (JSON)
Version
v1Published
First published
Evidence type
Controlled or validated comparison
Sources
1 sourcePeer-reviewed
sha256
4e540ed10aac…check it

Giving empathy is good, but providing emotional support is hard work. Empathic AI gives cheap and easy support, but may be ethically costly. When AI provides empathy, people may feel less compelled to support others. Four studies provide evidence for “empathy offloading”: when people see AI as capable of empathy, they become less willing to empathize with the suffering of others, even without realizing it. When AI seems more capable of empathy, people—both with correlations (Study 1) and causal experiments (Stud…

The problem

In four studies, people who saw AI as capable of empathy were less willing to empathize with others' suffering, and learning that someone used an LLM for support led to fewer empathic responses and more social judgment.

What this doesn’t fix

The authors describe empathic devaluing and atrophy, moral deskilling and societal fraying as possible implications, not outcomes the four studies measured.

Sources

1 on file
  1. Computers in Human Behaviordoi:10.1016/j.chb.2026.109240
    Peer-reviewed2026-10-28

Cite this record

v1

Truvace Impact Record TRV-2026-2502, v1: “People offload empathy when they think AI can give it instead.” Truvace, 2026-11-12. /record/TRV-2026-2502 (accessed on the day you copy it). sha256 4e540ed10aac53e9…

Change history

1 version

Every change made to this record since it was first published, and why. Nothing has changed since it went up.

  1. Publishedv14e540ed10aac…

    Certified into the record

Check this record hasn’t been altered

Check this record hasn’t been altered
How to check it yourself, without trusting this page

Download any version from /api/record/TRV-2026-2502 and hash its canonical field yourself, for example with shasum -a 256. You should get its content_hash. Each version’s text also ends with prev:and the hash of the version before it (version 1 points to 64 zeros). If anyone had changed a single character of any version after it was saved, those hashes wouldn’t line up.