Department of Psychology, University of Warwick, Coventry, England.
Psychol Rev. 2012 Oct;119(4):888-92; discussion 893-8. doi: 10.1037/a0028004.
Gonzalez and Dutt (2011) recently reported that trends during sampling, prior to a consequential risky decision, reveal a gradual movement from exploration to exploitation. That is, even when search imposes no immediate costs, people adopt the same pattern manifest in costly search: early exploration followed by later exploitation. From this isomorphism in costless and costly search, the authors concluded that the same cognitive mechanisms underlay the control of sampling in 2 experimental paradigms employed to investigate decisions from experience: the sampling paradigm implementing costless search and the repeated-choice paradigm implementing costly search. We show that this is a misinterpretation of the human data resulting from drawing inferences about cognitive processes from data aggregated across individuals. Because of an inverse relationship between sample size and the propensity to explore, aggregating across individuals produces a pattern where exploration is gradually replaced by exploitation. On an individual level, however, there is no general reduction in exploration in the sampling paradigm. We list ensuing problems for the instance-based learning model Gonzalez and Dutt presented to explain the similarities between sampling and repeated decisions from experience.
冈萨雷斯和达特（2011 年）最近报告称，在进行重要风险决策之前的采样过程中，趋势显示出逐渐从探索到开发的转变。也就是说，即使搜索不立即产生成本，人们也会采用与昂贵搜索中相同的模式：早期探索，然后是后期开发。从无成本和有成本搜索的这种同构性，作者得出结论，在用于研究经验决策的两个实验范式中，相同的认知机制控制了采样：实施无成本搜索的采样范式和实施有成本搜索的重复选择范式。我们表明，这是对人类数据的误解，因为从个体数据中推断认知过程会导致误解。由于样本大小与探索倾向之间的反比关系，跨个体进行聚合会产生探索逐渐被开发取代的模式。然而，在个体层面上，采样范式中并没有普遍减少探索。我们列出了冈萨雷斯和达特为解释采样和重复经验决策之间的相似性而提出的实例学习模型所面临的后续问题。
