Evidence
Which observations help distinguish competing explanations? What do they establish, and what remains unresolved?
How understanding becomes action.
Eidos is a research and technology company exploring how people and institutions make decisions under uncertainty.
We study the relationship between evidence, judgment, and action: how beliefs form, how they change, and when they provide a basis for a decision.
Inquiry
What gives a decision its foundation?
Every decision rests on a view of the world. We are interested in how that view is formed, tested, and revised.
Judgment
How do people weigh incomplete information? What can disagreement reveal about assumptions and uncertainty?
Action
When is the evidence sufficient to act? How should the consequences of being wrong shape that choice?
Learning
What happens when expectations meet outcomes? How can institutions recognize errors and revise their understanding?
An experiment in judgment
What would change your mind?
A signal is appearing in a sequence of observations. Is there a persistent pattern, or is it background variation? Reveal evidence and watch the balance shift.
Explanation A
A persistent pattern
A signal appears in 75% of observations.
Explanation B
Background variation
A signal appears in 25% of observations.
Observations
No observations yet. Both explanations start equally plausible.
Pattern 50%Variation 50%
The evidence has not yet favored either explanation.
The same evidence can lead to different decisions when the consequences differ. For institutions, weighing evidence is only part of deciding what to do.
About this experiment
This is an illustrative model with two possible explanations and equal starting probabilities. Each observation is assumed independent. A signal is three times as likely under the pattern explanation; its absence is three times as likely under background variation. Bayes’ rule updates the balance after each observation.
The reveal button provides eight observations in a fixed sequence. You can add conflicting evidence up to a total of twelve observations. Conflicting evidence adds an observation favoring the currently less likely explanation. At an even balance, it adds an absence. These are constructed examples, not collected data or a measure of real-world confidence. Other explanations and dependent observations would change the model.
Perspective
More information can leave us less certain.
A new observation can strengthen an explanation, expose a missing assumption, or reopen a question we thought was settled.
Make assumptions visible.
Understanding a conclusion means understanding what it depends on, including the possibilities that have been left out.
Give disagreement a purpose.
Different judgments can reveal different evidence, different expectations, or different consequences of being wrong.
Leave room for revision.
A decision can be necessary before a question is settled. Learning requires keeping track of what would make us reconsider.
Contact
Continue the conversation.
We welcome conversations about evidence, judgment, and how institutions learn.