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-2026-2698
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-12T17:02:18.219186Z
status: published
lens: g_space
sector: business
headline: Modelling user trust in chatbots: the role of anthropomorphic cues and social responses in information services
dek: Introduction. Chatbots, as an emerging form of information technology, are increasingly integrated into information services worldwide, yet how their anthropomorphic design features shape users’ psychosocial responses remains underexplored. This study builds on the Computers Are Social Actors (CASA) paradigm and examines the effects of chatbots’ anthropomorphic cues on users’ perceived social presence, parasocial interaction, and trust in chatbots. Method. A between-subjects experiment was conducted with 195 par…
gain_title: In a randomized experiment with 195 Chinese adults reading a mock emotional-support chat, those shown a human-like chatbot rated its social presence higher (6.00 vs 5.61 on a 7-point scale), and that presence accounted for an indirect rise in their reported trust (indirect effect 0.254).
problem_title: (none)
trace_subject: (none)
gain_reading: In a randomized experiment with 195 Chinese adults reading a mock emotional-support chat, those shown a human-like chatbot rated its social presence higher (6.00 vs 5.61 on a 7-point scale), and that presence accounted for an indirect rise in their reported trust (indirect effect 0.254).
gain_evidence: Participants in the human-like chatbot group reported higher levels of social presence | Social presence served as a key mediator linking anthropomorphism to trust
problem_reading: (none)
problem_evidence: (none)
quick_read: A between-subjects experiment with 195 Chinese adults found that a chatbot given human-like cues was rated as more socially present than a machine-like one. Participants who saw the human-like version scored social presence at 6.00 on a 7-point scale, against 5.61 for the machine-like version.

Anthropomorphism did not raise trust directly. Its effect ran through social presence: the indirect path to trust was 0.254 (95% CI 0.083 to 0.462). Parasocial interaction, the feeling of a one-sided relationship with the bot, did not significantly affect trust.

The results come from self-reported ratings after reading a scripted emotional-support chat, not from live conversations. The authors say real-world interactions, other age groups and task-based settings such as customer service still need testing.
limitation: Participants only read a scripted, one-shot chat mock-up rather than talking to a live chatbot, and the sample skewed young and female, so the trust effect may not carry over to real, sustained use.
tag: Evidence-backed gain
key_points: Participants were Chinese adults aged 19 or older with chatbot experience, recruited on the Credamo platform; 203 were recruited and 195 valid responses were analysed (97 human-like, 98 machine-like). | The human-like chatbot introduced itself as Xiaoyi with a human avatar, warm personalised replies and emoji; the machine-like one used a robot avatar and brief impersonal answers. | Social presence was higher with the human-like chatbot (M = 6.00, SD = 0.64) than the machine-like one (M = 5.61, SD = 1.19); b = 0.395, p = 0.004. | Anthropomorphism had no significant direct effect on parasocial interaction or on trust (both p > 0.05). | Social presence strongly predicted trust (b = 0.643, p < 0.001) and parasocial interaction (b = 0.962, p < 0.001). | The indirect path from anthropomorphism to trust through social presence was significant (0.254); the path through parasocial interaction (-0.009) and the chained path (0.024) were not. | The authors advise against blind trust, noting that users should be cautious with chatbot advice in high-stakes settings such as health questions.
rundown: Participants read a short scenario asking them to imagine feeling low and seeking anxiety relief from a chatbot on a website, then viewed one of two mocked-up chat screens. They had to read the material for at least one minute before completing seven-point Likert questionnaires on social presence, parasocial interaction and trust. A pre-test with 25 people and a manipulation check in the main sample confirmed that the human-like version was perceived as more human-like.

The analysis used SPSS, AMOS and the PROCESS macro with 5,000 bootstrap samples. The scales showed good reliability (Cronbach's alpha 0.907 for social presence, 0.902 for parasocial interaction, 0.859 for trust). Overall, participants rated social presence at 5.81, trust at 5.65 and parasocial interaction at 5.36 on average.

Two hypotheses about direct effects of human-like cues on parasocial interaction and trust were not supported, and neither was parasocial interaction as a mediator. The authors suggest parasocial interaction may need repeated exposure to build into a lasting bond, and that trust may also depend on competence, credibility and attractiveness.

Based on the findings, the authors recommend that developers of emotional-support and information chatbots add emotionally responsive language, human-like avatars and personalised messages. They also call for transparency about information sources and limits, and for accountability rules so organisations cannot exploit user trust.
sources:
- peer_reviewed | Information Research an international electronic journal | https://doi.org/10.47989/ir31261439 | 2026-05-15
prev: 0000000000000000000000000000000000000000000000000000000000000000

SHA-256 hash

567e324f764121176c350bcfeceecc142019e769c7acf052a0f590123117c92e

Previous version’s hash

0000000000000000000000000000000000000000000000000000000000000000