社会科学预测中的准确率、偏差与虚假确定性
Forecasting a social experiment is a demanding test of judgment. It requires translating theory, context and prior evidence into a numerical expectation about how people will respond.
A recent study reports that large language models estimated experimental results about as accurately as groups of human forecasters, including for studies published after the models’ training period. However, the systems tended to overstate the size of effects.
Inspired by a public Nature title/summary · Rewritten by Vocabsavvy · Vocabsavvy Original (public-snippet inspired)