Reflexive Knowledge Model
Update forecasts when knowledge changes the behavior being forecast
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 94%
Knowledge about an adaptive social system can change the system itself. A credible forecast influences the decisions of investors, consumers, competitors, or policymakers; those decisions modify present prices and conditions, invalidating the original inputs. Harari illustrates the mechanism with an oil-price prediction: if people accept a forecast of a future increase, buying and repricing begin immediately, changing the data from which the forecast was produced. The practical response is not to abandon forecasting but to make it reflexive. Analysts must model who receives the knowledge, how they are likely to react, and what second-order effects those reactions create. Forecasts should therefore be treated as changing scenarios that require revision after dissemination.
Origin
Extracted from Masters in Business through Harari's explanation of why behavior-changing historical or market knowledge rapidly loses relevance.
Core principles
- 01Useful knowledge changes behavior.
- 02Behavioral change alters the system being observed.
- 03Public predictions can help invalidate themselves.
- 04Forecasts about adaptive systems require continuous updating.
- 05The impact of a prediction is part of the prediction problem.
How to run it
- 1
Define the forecast
State the expected outcome, time horizon, assumptions, and affected actors clearly.
Pro tip Separate observed facts from behavioral assumptions.
Watch out A vague prediction cannot be tested or updated coherently.
- 2
Map informed actors
Identify who may receive, believe, reject, or act on the prediction.
Pro tip Include regulators, competitors, suppliers, and customers as well as direct decision-makers.
Watch out Do not assume all actors respond identically.
- 3
Model immediate reactions
Estimate how informed participants could change purchases, prices, policies, or resource allocation before the predicted event.
Pro tip Look for actions that can be taken sooner than the forecast horizon.
Watch out Ignoring anticipation effects produces stale projections.
- 4
Recalculate the system
Update the original assumptions using the changed present conditions created by those reactions.
Pro tip Maintain several scenarios for different adoption levels.
Watch out Do not preserve the original forecast merely to appear consistent.
- 5
Repeat as conditions evolve
Monitor responses and revise the prediction as new rounds of behavior alter the system.
Pro tip Set explicit triggers for reassessment.
Watch out Adaptive systems may never settle into a single final forecast.
In the wild
A credible prediction says oil will cost one hundred dollars more in a year. Market participants act on that knowledge immediately, pushing today's price toward the forecast level. The new price changes supply, demand, incentives, and calculations, leaving the original one-year estimate outdated.
→ The prediction changes the present and makes the future diverge from the initially forecast path.
A retailer forecasts that a rival will cut prices next quarter and shares the analysis internally. Buyers renegotiate supply contracts and managers launch promotions early. The rival observes these moves and changes its plan, so the team updates its scenarios rather than adhering to the original forecast.
→ The strategy remains responsive to second-order competitive behavior.
Common mistakes
Treating people like passive variables
Participants who learn a forecast may adapt strategically rather than continuing their previous behavior.
Ignoring publication effects
A private forecast and a widely publicized forecast can produce radically different outcomes.
Defending an obsolete prediction
Behavior-changing knowledge requires recalculation; consistency is not accuracy when the underlying system has changed.
Is it for you?
Best for
It is best for market forecasts, competitive strategy, public policy, and other domains where participants react to information.
Not ideal for
It is not ideal for stable physical processes whose behavior is unaffected by people learning the forecast.
From the transcript
“Because once the knowledge you have changes the way in which you behave, um, the world changes and your knowledge is no longer updated.”
“This is how predictions about the future change the present, and then the future becomes different from what you predicted.”
From the episode
Interview With Yuval Noah Harari: Masters in Business (Audio)
Yuval Noah Harari