On 18 October 2013, Émile Servan-Schreiber announced a video of his École Militaire lecture on collective intelligence and geopolitics. A contemporary report by Loïc Salmon describes the event’s date and central argument. The recording is unavailable, so the lecture themes below follow Salmon’s account.
A defence audience, a forecasting problem
Writing for the Association nationale des croix de guerre et de la valeur militaire, Salmon dates the conference-debate to 8 October 2013 in Paris. It identifies Servan-Schreiber as managing director of Lumenogic and the organiser as the Association nationale des auditeurs jeunes de l’Institut des hautes études de défense nationale, or ANAJ-IHEDN.
The report presents his argument as a response to dispersed knowledge: geopolitical developments draw on information that no individual can possess in full. Diverse, independent assessments and information from people close to events could help reduce uncertainty. That was an argument about how to organise judgment, rather than a claim that a crowd automatically knows the future.
Markets make disagreement measurable
In Salmon’s account, prediction markets give participants an incentive to find useful information and revise their positions. Reputation can be at stake even where money is absent. The report also distinguishes a continuously updated forecast of an eventual election result from a poll recording preferences at a particular moment. In that distinction, a statement about a future outcome and a measurement of present preferences serve different purposes.
The account connects this approach to the Good Judgment research effort and Lumenogic. The institutional context is independently documented by IARPA’s Aggregative Contingent Estimation programme: its stated goal was to improve intelligence forecasts by eliciting, weighting and combining judgments, with accuracy tested against real events. That programme design gives a practical way to test the promise of collective forecasting: ask defined questions, combine the estimates and compare them with what happens.
Testing collective forecasts against events
A later primary record supplies a concrete research outcome. On 28 April 2017, IARPA announced the public release of Good Judgment data containing millions of forecasts across four years of competition. That release concerns the broader research programme, not an evaluation of this Paris lecture.
The journal’s history of NewsFutures and superforecasting follows that research connection further. The 2013 event captures an earlier moment when collective forecasting was being explained to a defence audience. Its methodological point is to turn disagreements about future events into dated judgements that can later be checked. Decentralised knowledge supplies inputs; a common question and an eventual result give the group a way to evaluate what those inputs achieved.
