正解: BCorrect: B
Tetlock の研究では、単一の理論に固執する専門家より、複数の視点を組み合わせて確率的に考える専門家の方が一貫して予測精度が高いことが示されました。Tetlock found that experts who combine multiple perspectives and think probabilistically consistently outforecast those who cling to a single grand theory.
Tetlock は20年にわたり約280人の専門家による数万件の政治・経済予測を追跡し、専門家全体の予測精度が単純な外挿と大差ないこと、そして「有名な専門家ほど精度が低い」傾向を発見しました。重要なのは、精度を分けたのが専門分野や学歴ではなく思考スタイルだった点です。Isaiah Berlin の比喩を借り、一つの大きな理論を信奉する「ハリネズミ」より、多くの小さな知識を柔軟に使う「キツネ」が優れていました。その後の Good Judgment Project では、訓練を受けた一般人チームが情報機関のアナリストを上回る成績を出しました。選択肢Aは事実と逆で、Cは専門知識の価値を全否定しており誤り、Dは学歴と精度の相関が弱いという知見に反します。Over two decades, Tetlock tracked tens of thousands of political and economic forecasts by roughly 280 experts and found that experts as a group barely beat simple extrapolation, and that more famous experts tended to be less accurate. Crucially, what separated forecasters was thinking style rather than field or credentials. Borrowing Isaiah Berlin's metaphor, 'foxes' who draw flexibly on many small pieces of knowledge outperformed 'hedgehogs' devoted to one big theory. The later Good Judgment Project showed that trained teams of ordinary volunteers could outperform intelligence analysts. Option A is the reverse of the finding; confident pronouncements play well in media but forecast poorly. Option C wrongly dismisses expertise, which remains essential for knowing what questions to ask. Option D contradicts the weak link between credentials and accuracy.
「専門家も当てにならない」という知見は、専門知識そのものへの不信や、根拠のない素人意見と専門家の意見を同列に扱う風潮を助長する危険があります。またキツネ型の専門家は断定を避けるため、迅速な決断が求められる場面では頼りなく見え、組織で影響力を持ちにくいという現実もあります。The finding that experts forecast poorly can fuel distrust of expertise itself and a tendency to treat baseless lay opinions as equal to informed ones. Fox-like experts avoid bold assertions, which can make them seem unreliable when quick decisions are needed and limit their influence in organizations that reward decisiveness.
専門家に意見を求める際は「何が起きるか」ではなく「どのような条件でどの確率で起きるか」「その予測が外れるとしたらどんな理由か」を尋ね、複数の専門家の見解を集約します。予測の的中率を記録して、断定の強さではなく実績で信頼度を判断する仕組みを作ります。When consulting experts, ask not 'what will happen' but 'under what conditions, with what probability, and what would make you wrong', and aggregate views from several experts. Keep a record of forecast accuracy so that trust is calibrated by track record rather than by the confidence of the delivery. Reward well-reasoned probabilistic judgments even when they turn out wrong.