Maverick Top Gun
Hβ2.176 readers tagged 5 emotions. Most landed on Proud (50%), with Surprised (50%) close behind. AI read it as Surprised β chosen by only 50% of human readers.
See full breakdown β
Live data Β· updating as readers join
The dataset, growing in real time.
Anonymized aggregate readings from MindLens Lab β a research project capturing how people actually read emotion in short social moments, so the data emotion AI trains on can line up with the phenomenon. Each new participant joins the totals below; new constellations appear once a clip clears 5 responses.
14
Participants
150
Responses
33
Curated clips
2
Countries
Reading from
- SG
- US
Constellation gallery
Where readers split β visualized.
Each card below is one clip. The night-sky shows how readers split across the emotions.
?What am I looking at?expand
A constellation here is a way of showing that the same moment can be read in many different ways at once β not as noise to clean up, but as the actual finding.
Each star is one of the emotions. Its size and brightness scale with how many readers picked that emotion: bigger and brighter = more readers chose it. Empty/dim stars are options that no one (or almost no one) picked.
The peach-glow star is the most-picked emotion β the modal reading. The mint-glow star is how one AI (Claude) read the same moment. When peach and mint sit far apart, the AI saw something humans largely didn't β that's a finding worth reading.
The Intern - Jules Apologizes
Hβ2.166 readers tagged 5 emotions. Most landed on Sad / sorry (50%), with Anxious / nervous (50%) close behind. AI read it as Embarrassed / awkward β chosen by only 50% of human readers.
See full breakdown β
Ryan Gosling and Harrison Ford Interview
Hβ2.166 readers tagged 6 emotions. Most landed on Happy / amused (83%), with Disappointed (17%) close behind. AI read it as Happy / amused β same as the human modal.
See full breakdown β
Love Actually
Hβ2.127 readers tagged 5 emotions. Most landed on Disappointed (57%), with Sad / sorry (43%) close behind. AI read it as Disappointed β same as the human modal.
See full breakdown β
When MJ Finds Out
Hβ2.058 readers tagged 5 emotions. Most landed on Surprised (63%), with Confused (38%) close behind. AI read it as Proud β chosen by only 13% of human readers.
See full breakdown β
The Devil Wears Prada
Hβ1.925 readers tagged 4 emotions. Most landed on Scared / afraid (40%), with Embarrassed / awkward (40%) close behind. AI read it as Embarrassed / awkward β chosen by only 40% of human readers.
See full breakdown β
Cards are sorted by Shannon entropy (H) β higher = more plural reading. Only clips with β₯ 5 responses appear; new constellations join as more readers participate.
AI vs human readings
Where the AI saw it differently.
On these clips, one AI (Claude) read the moment as an emotion that few human readers agreed with. The divergence isn't the AI being βwrongβ β in a plural-reading framework, no single answer is right. But it does suggest that AI and humans are weighting different cues, and that's the question the project is trying to map.
βWhen MJ Finds Outβ
n=8- Surprised5 Β· 63% (3p)
- Confused3 Β· 38% (1p)
- Happy / amused2 Β· 25% (1p)
- Proud1 Β· 13% (0p)
- Anxious / nervous1 Β· 13%
63% of readers landed on Surprised.
AI saw Proud β no one picked it as their primary, but 13% tagged it as part of their Mixed reading.
See full breakdown β
βShrinking - TV dramaβ
n=6- Embarrassed / awkward5 Β· 83% (4p)
- Affectionate / tender2 Β· 33% (1p)
- Happy / amused1 Β· 17% (0p)
- Anxious / nervous1 Β· 17% (0p)
83% of readers landed on Embarrassed / awkward.
AI saw Happy / amused β no one picked it as their primary, but 17% tagged it as part of their Mixed reading.
See full breakdown β
βModern Loveβ
n=6- Sad / sorry5 Β· 83% (4p)
- Moved / touched1 Β· 17%
- Affectionate / tender1 Β· 17% (0p)
83% of readers landed on Sad / sorry.
AI saw Moved / touched β tagged by 17% of readers in some form.
See full breakdown β
βIron Man Takes Spiderβ
n=6- Angry / frustrated6 Β· 100% (5p)
- Disappointed1 Β· 17% (0p)
- Contempt / scorn1 Β· 17% (0p)
100% of readers landed on Angry / frustrated.
AI saw Disappointed β no one picked it as their primary, but 17% tagged it as part of their Mixed reading.
See full breakdown β
Ranked by how few humans agreed with the AI's reading β lowest agreement first. This is exactly the kind of pattern that a single-answer AI tool would smooth over and a plural dataset can surface.
About the methodology
Each reading captures two axes (emotion + cues) and is measured for plurality with Shannon entropy and Krippendorff's Ξ±. The full taxonomy, metrics, and pre-registered hypotheses live on the research pages.
Reader diversity
Who's contributing
Optional details participants choose to share. Each card appears once at least 5 participants have answered that question. Distributions are descriptive β they show the texture of the dataset, not statistical findings.
Primary language
Shared by 5 of 14 participants
- Korean (νκ΅μ΄)3 Β· 60%
- English2 Β· 40%
Confidence in English
Shared by 7 of 14 participants
- 45 Β· 71%
- 52 Β· 29%
By collection
Per-collection summaries
Collection 3
10 participants Β· 2 countries Β· 70 responses Β· open
Most varied reading
On βMaverick Top Gunβ, readers tagged 5 different emotions across 6 readings.
- Proud3 Β· 50% (1p)
- Surprised3 Β· 50% (1p)
- Anxious / nervous2 Β· 33% (1p)
- Scared / afraid1 Β· 17% (0p)
- Confused1 Β· 17% (0p)
Where we are
A multi-year project, growing alongside its researcher
Phase 1
Plural reading dataset
Phase 2
Validated study materials
Phase 3
Adaptive tools
See the full timeline β
Cite this dataset
Treat MindLens Lab as a citable academic asset.
APA
Kim, E. (2026). MindLens Lab: Plural Emotion Reading Dataset (Phase 1) [Dataset]. https://mindlenslab.org
BibTeXshow
@dataset{kim2026mindlens,
author = {Kim, Evelyn},
title = {MindLens Lab: Plural Emotion Reading Dataset (Phase 1)},
year = {2026},
version = {1.0 β rolling},
url = {https://mindlenslab.org},
note = {Anonymous, ongoing collection of plural human readings of social-emotional video clips, paired with cue annotations and one AI's reading of the same clips.}
}Phase 1 is an ongoing project β the dataset version above increments with each closed collection. For the full anonymized export (responses, distributions, AI annotations), email contact@mindlenslab.org.
Add your reading
You've seen the spread. Help shape it.
Every constellation above is built from real participants' readings. Yours would join the same dataset β short clips, 10 to 15 minutes, no right or wrong answers.
Methodology + transparency
Each row above is one anonymous response. Counts include only responses kept in analysis (excluded responses are removed). Internal test profiles are excluded from every aggregate.
Want the full anonymized dataset? Email us.