The exact text we saved

This is the version exactly as it was saved, character for character. The SHA-256 hash below is a 64-character code worked out from that text. Run the same text through SHA-256 and you get the same code; change a single letter and the code comes out completely different. Each version also stores the previous version’s hash, which chains them together so no version can be swapped out quietly.

TRUVACE RECORD VERSION
record: TRV-2026-2548
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-12T07:51:59.502742Z
status: published
lens: g_space
sector: policy
headline: How emotional response styles of generative AI companions shape adaptive emotion regulation and short-term AI reliance
dek: Introduction Generative artificial intelligence (GenAI) companions increasingly respond to users' negative emotions with validation, reassurance, reframing, and personalized guidance. However, immediate emotional relief does not establish improved independent emotion regulation or emotional dependence. Methods In a two-condition randomized experiment with 380 adults, we compared adaptive regulation-oriented responses with relational reassurance-oriented responses. We assessed immediate negative affect, emotion-r…
gain_title: In a randomized trial of 380 adults, AI replies that coached emotion regulation led to higher self-efficacy and better affective recovery in a later no-AI task (0.99 vs 0.71 points) than reassurance-style replies.
problem_title: (none)
trace_subject: (none)
gain_reading: In a randomized trial of 380 adults, AI replies that coached emotion regulation led to higher self-efficacy and better affective recovery in a later no-AI task (0.99 vs 0.71 points) than reassurance-style replies.
gain_evidence: Adaptive regulation-oriented responses were associated with higher emotion-regulation self-efficacy ( b = 0.500, p < 0.001) | mean no-AI affective recovery was 0.99 vs. 0.71
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers in Guangzhou randomly assigned 380 adults to talk through a mildly upsetting memory with an AI companion for three turns. One version coached them to label the emotion, reframe it and pick a coping step; the other soothed them and stressed it was always available. Both cut negative affect by about 1.6 points on the scale, with no significant difference in immediate relief.

The differences came afterwards. The coaching group reported higher self-efficacy (b = 0.500), had better blind-coded written strategies (b = 0.353) and recovered more during a later task without AI (0.99 vs 0.71 points). The reassurance group more strongly believed they needed the AI for comparable relief (b = 1.010) and showed stronger intentions to rely on it (b = 0.628).

The results suggest that judging AI companions only by how much better users feel in the moment misses differences in what users take away. The authors stress the reliance findings are short-term beliefs and intentions, not emotional dependence. In an exploratory choice, 36.8% of the reassurance group asked for another AI reply versus 27.9% of the coaching group, a difference that was not statistically conclusive (p = 0.063).
limitation: The study used a single three-turn interaction with mild non-clinical material, adults in Guangzhou, one text-only model snapshot and a written lab task, so it cannot establish emotional dependence or real-world coping; the AI-necessity and reliance scales were study-specific and need independent validation.
tag: Evidence-backed gain
key_points: Participants were 380 adults recruited in Guangzhou (190 per condition), with a mean age of 25.2; 396 completed the session and 16 were excluded under preset rules. | The companion ran on OpenAI's GPT-5.5 through its API, with three AI turns and three participant turns, in a Chinese-language interface that disclosed it was an AI. | Immediate relief was the same: negative affect fell 1.67 points with coaching and 1.61 with reassurance (interaction p = 0.323). | In a later no-AI task, negative affect after unaided regulation was 3.49 with coaching and 3.77 with reassurance, from 4.48 in both groups beforehand. | Blind raters scored written coping strategies higher after coaching (b = 0.353, p = 0.009); inter-rater reliability was ICC = 0.87. | Removing the items closest to the manipulation reduced the self-report effects by 26.5% to 34.2% but did not eliminate them. | Indirect-effect estimates were 0.279 for self-efficacy and 0.375 for perceived AI necessity, but the authors say these do not establish causal mediation.
rundown: The experiment was a two-condition randomized online study with concealed permuted-block allocation. After a mild autobiographical emotion induction, each participant had three exchanges with a GPT-5.5-based companion given one of two system prompts. Pilot testing with 80 other adults confirmed the two styles were comparable in warmth, empathy and helpfulness, and transcript coding found 95.3% and 95.8% of replies met the full fidelity threshold.

About three minutes later, participants faced a second standardized setback with no AI and wrote how they would regulate their response. Affective recovery during that task was 0.99 points after coaching versus 0.71 after reassurance (adjusted b = 0.279). All five focal effects survived Holm correction, and fidelity-restricted and expanded-covariate analyses changed coefficients by at most 0.015.

The reassurance prompt told the AI to emphasize continued availability and encourage returning for comfort. It did not claim consciousness or exclusivity and did not discourage human support. Participants in that group rated the AI as more necessary and said they would rely on it more, yet in the exploratory request-another-AI choice the gap was inconclusive (OR = 1.51, 95% CI 0.98 to 2.32).

The authors say the findings point to a design goal of warm but bounded support: validation combined with emotion labeling, limited reappraisal and a user-chosen action, while scrutinizing response policies that repeatedly position the AI as the source of calm.
sources:
- peer_reviewed | Frontiers in Psychology | https://doi.org/10.3389/fpsyg.2026.1950163 | 2026-09-08
prev: 0000000000000000000000000000000000000000000000000000000000000000

SHA-256 hash

d45ad0acd94f38b410e3a42e5011932d1a8b0eff6a45997a37a0e44ac40c72e0

Previous version’s hash

0000000000000000000000000000000000000000000000000000000000000000