The explicitly taught representations showed 74.20% mean participant-level graded agreement with subsequent participant judgments.
That was 5.24 percentage points higher than our specified indirect-attribute baseline.
2
We’ve published an expanded version of our earlier research on computational representation of individual judgment.
Same 99 participants. Same 4,460 judgments.
More detailed analysis. More baselines.
The numbers are looking more interesting.
zenodo.org/records/23179591
1
3
OpenSim has a new home.
We’ve been building, experimenting, and researching how human decisions can be represented and simulated.
The new site brings together the research, experiments, and things we’re building.
opensim.io
10
Emotion, sentiment, belief, opinion, and judgment are related—but they are not the same thing.
Where do they sit within Judgment = Disposition × Context × Constraints?
And how much can each tell us about how someone will judge a situation?
1
17
Emotion belongs primarily in context, not disposition.
It may temporarily reweight existing preferences without rewriting them.
But capturing it reliably—without friction, ambiguity, intrusion, or false confidence—is hard.
Reserved in the model. Deferred in implementation.
1
19
New research note:
Context and Preference Salience
Early observations on how explicit context changes simulated food choices.
PDF ↓
opensim.io/Context_Injection…
17
We observed something interesting:
- when choices were already close, context often tipped the outcome.
- when preferences were strong, context shifted distributions without replacing them.
16
This one surprised us.
We thought Croque Madame would win.
It didn’t.
Try it: menusim.io/play/5xR3zAMXUsKu…
18
We built a way to simulate how people choose.
Not predictions. Not surveys.
Just modeled decision behavior.
Today we’re releasing:
– a technical paper
– MenuSim, a live simulation playground
Full write-up + demos: medium.com/@opensimlabs/we-b…
Here’s one example ↓


