Six data points on how fast, and how reliably, we size up someone we've just met, from split-second face and voice judgments to the AI faces now gaming that instinct
The Trust-in-Seconds Dataset & Dashboard, DT Linux
The Trust-in-Seconds Dataset & Dashboard
An interactive companion to DT Linux's briefing on how fast, and how reliably, people judge a stranger's trustworthiness: split-second face and voice judgments, two popular explanations that don't hold up under review, and the modern systems now gaming the same instinct.
Filter by case
This report's three mechanisms
How each case compares
What's in this dataset, by composition
How well people spotted an AI-generated face, before any training
A genuine part-to-whole composition from case six's first test: every judgment was either correct or it wasn't. This is the one place in this dataset where the numbers are shares of a single total; every other comparison in this dashboard is an independent-category comparison and stays a bar chart on purpose.
Dataset schema
| Field | Description | Example |
|---|---|---|
| case_name | The judgment or mechanism this row describes | "The eye contact sweet spot" |
| tier | This report's own mechanism categorization, not a strength ranking | "The fine print" |
| key_metric | The headline sourced figure for this case | "3.2 second average comfortable gaze" |
| source | Citation for the key metric | "Royal Society Open Science, 2016" |
| last_verified | Date the figure was last checked against its original source | "2026-09-03" |
About this dataset
How this data was collected
Every figure behind this dataset comes from a peer-reviewed psychology study, a published critical review or replication attempt, or documented platform and industry data: Willis and Todorov's face-perception research and the meta-analysis that followed it, a voice-only trust modelling study, a Royal Society Open Science gaze-duration study, a critical review and a failed replication on oxytocin and trust, platform reputation research summarized in the Annual Review of Economics, and a 2022 PNAS study on AI-generated faces. DT Linux's research process cross-references each figure against its original publication; this report does not involve DT Linux collecting behavioral data itself.
Data handling and compliance
Built entirely from published academic research and publicly reported platform statistics, never from personal data on identifiable individuals collected by DT Linux. Full practices in our privacy policy.
API and pipeline access
Want a literature synthesis or trust and reputation-data analysis like this refreshed on a schedule, or piped into your own systems? DT Linux builds custom research and data pipeline work per client; there is no self-serve API today, so delivery is handled directly by our team. Get in touch.
More from the DT Linux data store
Want this kind of behavioral or trust research built for your own sector?
DT Linux designs custom literature syntheses, survey research, and trust or reputation-data analysis for organizations that need to understand the psychology and data behind their own customers, users, or hiring process. Get in touch if you need deeper or more advanced research behind this report.