Investigating cognitive placebo effects in reinforcement learning:A computational modelling approach

研究强化学习中的认知安慰剂效应：一种计算建模方法

• 批准号: 392135994
• 负责人: Dr. Zsolt Turi, Ph.D.
• 金额: --
• 依托单位: Tagesklinik für kognitive Neurologie
• 依托单位国家: 德国
• 项目类别: Research Fellowships
• 财政年份: 2017
• 资助国家: 德国
• 起止时间: 2016-12-31 至 2017-12-31
• 项目状态: 已结题
• 来源: https://gepris.dfg.de/gepris/projekt/392135994?language=en
• 关键词: Investigating; cognitive; placebo; effects; reinforcement

Placebos and nocebos are physiologically inert substances (e.g. pharmaceuticals) or simulated interventions (e.g. medical devices), which produce complex psychobiological responses in the participants despite the fact that they do not have any direct therapeutic effects. Whereas the term "placebo" describes positive effects (i.e., improvement in symptoms), "nocebo" relates to any negative effects (i.e., worsening of symptoms) of a simulated treatment. So far, the fundamental paradigms of placebo/nocebo effects have been primarily confined to the pain or motor domains and our knowledge about cognitive placebo effects remains poorly understood. In this research fellowship, we will ask the questions (1) whether verbal suggestions in conjunction with active sham protocols of transcranial direct current stimulation (tDCS) can induce cognitive placebo effects in different forms of reward learning and (2) characterize the role of uncertainty-based verbal suggestions in inducing cognitive placebo effects. To address these questions, we will focus on reward learning by using a computational modeling approach called reinforcement learning. We will concentrate on two forms of control systems of reinforcement learning that are called model-free (MF) and model-based (MB) systems. In order to dissociate between the MF and MB control systems, we will employ a well-characterized sequential two-choice Markov decision task. To induce cognitive placebo effects, we will use active sham protocols of tDCS in conjunctions with the systematic manipulation of uncertainty about treatment efficacy by employing verbal suggestions. To better understand the relationship between uncertainty about treatment efficacy and placebo response, participant will be randomly allocated into one control and three placebo manipulation groups, namely low-, mid- and high-uncertainty groups. The present research fellowship will contribute to the understanding how cognitive placebo effects are generated and maintained, which is a fundamental question in basic neuroscience research and it has important implications in clinical practice. Furthermore, this research fellowship will greatly extend my expertise in Bayesian statistical analysis and in computational modeling of reinforcement learning.

安慰剂和反安慰剂是生理惰性物质（如药物）或模拟干预措施（如医疗设备），尽管它们没有任何直接的治疗效果，但它们会在参与者中产生复杂的心理生物学反应。术语“安慰剂”描述的是积极效果（即症状的改善），而“反安慰剂”涉及模拟治疗的任何负面效果（即症状的恶化）。到目前为止，安慰剂/反安慰剂效应的基本范式主要局限于疼痛或运动领域，我们对认知安慰剂效应的了解仍然很少。在本研究中，我们将提出以下问题：(1)言语暗示结合经颅直流电刺激（tDCS）的主动假方案是否能在不同形式的奖励学习中诱导认知安慰剂效应；(2)表征基于不确定性的言语暗示在诱导认知安慰剂效应中的作用。为了解决这些问题，我们将通过使用一种称为强化学习的计算建模方法来关注奖励学习。我们将集中讨论两种形式的强化学习控制系统，即无模型（MF）和基于模型（MB）的系统。为了分离MF和MB控制系统，我们将采用一个特征良好的顺序二选择马尔可夫决策任务。为了诱导认知安慰剂效应，我们将使用主动的假性tDCS方案，并通过使用口头建议来系统地操纵治疗效果的不确定性。为了更好地了解治疗疗效的不确定性与安慰剂反应之间的关系，参与者将被随机分为一个对照组和三个安慰剂操纵组，即低、中、高不确定性组。本研究将有助于理解认知安慰剂效应是如何产生和维持的，这是基础神经科学研究的一个基本问题，在临床实践中具有重要意义。此外，这个研究奖学金将极大地扩展我在贝叶斯统计分析和强化学习的计算建模方面的专业知识。