Skip to content
Kapari
How it works
Who it's for
Executives HR Agencies Consultants
Examples Proof Security Pricing Demo Version française Log in Start your 30-day free trial
The bench, taken apart

The decision theories Kapari runs on

Fifty years of research keep saying the same thing: we decide badly, in predictable ways, and that can be fixed. Here are the seven works that matter, what each one changes for you, and what Kapari does with it.

iEvery source is cited. Where Kapari applies only part of a theory, it says so.
What this is about

We decide badly, in predictable ways. That can be fixed.

Since the 1970s, research has converged: human judgment leans the same way, case after case (that is bias), and varies from one expert to the next on the same file (that is noise). The fix is not demanding perfect leaders. It is equipping the way you decide, by bringing disagreement into the open before the announcement.

That is what Kapari does: you describe a decision, simulated voices, built from public data, react to it, and you see the objections before you commit. Across 100 real decisions from the past, the engine found 77% of the objections documented at the time. See the results.

1 · Kahneman, Lovallo and Sibony, 2019

You will be judged on how you decided.

A good decision can end badly, and a bad one can succeed: luck scrambles the grade. The only workable rule is to judge how the decision was made: who argued the other side, which uncertainties were explored. Those questions rarely get asked in a meeting, because disagreeing is costly.

In Kapari: at the end of each test, Kapari counts the real disagreements and the blind spots, meaning what no voice defended. These are counts made by the software.

The full story: A good decision can end badly. So what do you grade?

2 · Kahneman, Sibony and Sunstein, 2021

A conclusion that does not depend on anyone's mood.

Two serious experts, the same file, two conclusions: that is the noise in human judgment. The fix has been known since Meehl (1954) and Dawes (1979): a simple rule, applied consistently, holds its own against case-by-case expert judgment, because it does not get tired or moody.

In Kapari: the conclusion comes from published rules; the AI only gives the characters their voice.

The full story: The conclusion is computed, not written by an AI.

3 · Gary Klein, 2007

When everyone says yes, look for what could break.

The premortem fits in one sentence: "the project failed, tell us why." Once failure is taken as given, the team says what politeness was holding back. Klein published it in Harvard Business Review in 2007.

In Kapari: when agreement is too broad, Kapari runs that premortem on its own.

The full story: When everyone agrees, Kapari raises a flag.

4 · Kahneman and Klein, 2009

Your intuition is valuable, except on rare decisions.

Intuition works where you practice often and get quick feedback. It goes off the rails on the decisions a leader makes once in a career.

In Kapari: the rule applies to the tool itself. Kapari tells you when not to trust it: a topic outside its documented sources, missing context, a decision that is not really a decision.

The full story: Our tool tells you when not to trust it.

5 · Kahneman and Lovallo, 1993, then Flyvbjerg

Your case feels unique. Others have tried it before.

Instead of running through your own plan, you compare your decision with similar decisions already tried, and with what became of them. Flyvbjerg made it the standard method for big projects.

In Kapari: a sourced reference point places your decision among similar ones; when no reliable one exists, nothing is shown.

The full story: Your case is less special than you think.

6 · Philip Tetlock, 2005 and 2015

Many angles beat a single expert.

Twenty years of measuring expert judgment: people who know one big thing get it wrong more often than people who know many small things. What protects you from a misstep is a diversity of viewpoints.

In Kapari: the voices are picked to cover every camp: several dozen contrasting voices, depending on the decision, rather than three experts who agree with each other.

The full story: Why dozens of voices beat three experts.

7 · Argyle et al., 2023

Believable simulated voices, with known limits.

The founding paper shows that an AI, fed real profiles, can reproduce the typical answers of human groups. The work that followed measures the limits: unstable answers, smoothed-out opinions, gaps between groups.

In Kapari: the voices are there to spot the camps and the objections, never to produce an opinion percentage.

The full story: Are AI synthetic audiences accurate? And the map of the market: synthetic audiences, the guide.

For the curious: the works cited
Kahneman, D., Lovallo, D., & Sibony, O. (2019). A Structured Approach to Strategic Decisions. MIT Sloan Management Review, 60(3), 67-73.
Kahneman, D., Sibony, O., & Sunstein, C. R. (2021). Noise: A Flaw in Human Judgment.
Meehl, P. E. (1954). Clinical versus Statistical Prediction. University of Minnesota Press.
Dawes, R. M. (1979). The robust beauty of improper linear models in decision making. American Psychologist, 34(7), 571-582.
Grove, W. M., Zald, D. H., Lebow, B. S., Snitz, B. E., & Nelson, C. (2000). Clinical versus mechanical prediction: a meta-analysis. Psychological Assessment, 12(1), 19-30.
Klein, G. (2007). Performing a Project Premortem. Harvard Business Review, 85(9), 18-19.
Kahneman, D., & Klein, G. (2009). Conditions for Intuitive Expertise: A Failure to Disagree. American Psychologist, 64(6), 515-526.
Kahneman, D., & Lovallo, D. (1993). Timid Choices and Bold Forecasts: A Cognitive Perspective on Risk Taking. Management Science, 39(1), 17-31.
Flyvbjerg, B. (2006). From Nobel Prize to Project Management: Getting Risks Right. Project Management Journal, 37(3), 5-15.
Tetlock, P. E. (2005). Expert Political Judgment: How Good Is It? How Can We Know? Princeton University Press.
Tetlock, P. E., & Gardner, D. (2015). Superforecasting: The Art and Science of Prediction.
Argyle, L. P., Busby, E. C., Fulda, N., Gubler, J. R., Rytting, C., & Wingate, D. (2023). Out of One, Many: Using Language Models to Simulate Human Samples. Political Analysis, 31(3), 337-351.

Test your next announcement.

Try the demo on a business decision, no account needed: 60 seconds is enough to see if it clicks.