摘要: 证据积累模型(evidence accumulation models, EAM)是关于人类决策过程的主要认知模型之一，其假定决策者不断搜集信息并将信息整合成与决策有关的证据，当累积证据量达到某个阈值时做出决策并反应。虽然EAM在研究中得到广泛应用，甚至有研究者认为其已经达到了理论的高原期，但EAM的理论预设并未被严格检验。以漂移扩散模型(Drift diffusion model，DDM)这一最具有代表性的EAM计算模型为例，其存在五个基本预设：（1）证据积累的普适性、（2）证据积累的选择性、（3）证据以线性方式积累、（4）决策标准恒定、（5）决策与运动执行过程独立。回顾对这五个基本预设进行检验的实证研究，可以发现：尽管DDM被广泛应用于知觉决策、记忆和基于价值的决策任务，但研究者仅在知觉决策任务中对证据积累是否存在进行了验证，在其他任务中多为直接应用；证据积累的选择性目前较缺乏实证研究；证据以线性方式积累的预设得到了较多知觉决策实验数据的支持，但在基于价值的决策中其是否成立仍然存在争议；决策标准恒定的预设则存在较大争议，促进了新证据积累模型的产生；决策独立于反应执行的预设近年来受到关注，但较多实证研究质疑了这一预设。总之，对证据积累模型的预设进行验证的实证研究并不均衡，部分预设的实证证据有限，亟需更多的实证研究进行验证。同时，通过对EAM预设进行清晰表述并回顾其实证证据，本研究表明清晰地表述模型预设有助于全面而系统地检验模型，从而不断地推动模型的更新与理论的发展，以更好地理解人类认知过程。
Abstract: The evidence accumulation models (EAM) are one of the main cognitive models of human decision-making, which assumes that decision-makers continuously gather information and integrate it into evidence relevant to the decision, a decision is made once the accumulated evidence reached a predefined threshold. Because of the widespread of these models, some researchers stated that it has reached a theoretical plateau. However, the theoretical assumptions of EAM have not been rigorously examined. Taking the drift diffusion model (DDM), the most widely used EAM computational model, as an example, there are five underlying assumptions: (1) the universality of evidence accumulation, (2) the selectivity of evidence accumulation, (3) linear integration of evidence, (4) a constant decision criterion, and (5) decision-making is independent of motor execution. After reviewing empirical studies that directly tested these five assumptions, we found that DDM has been widely applied to perceptual decision-making, memory and value-based decision-making tasks, among others, but has only been verified the evidence accumulation process in perceptual decision-making tasks. The assumption that evidence accumulates linearly has been supported by data from perceptual decision-making experiment, but its validity in value-based decision-making is questioned. As for the assumption of the selectivity of evidence accumulation, there lacks of enough empirical studies. The fourth assumption, the decision criterion is constant, is highly controversial, several other evidence accumulation models challenged this assumption. The last assumption, decision-making is independent of motor execution, only received attention in recent years, but empirical studies questioned this assumption. Together, our review suggested that empirical evidence for the assumptions of DDM is not balanced, and some assumptions were only examined by few empirical studies. At the same time, by clearly stating the model assumptions and reviewing the empirical evidence, we showed that clear and transparent specifications of model assumptions are helpful for a comprehensive and systematic evaluation of the model, thus facilitate the update of models, the development of theories, and ultimately, deepen our understanding of human cognitive processes.