情緒活動對酬賞基礎決策歷程之影響

Abstract

我們的日常生活中充滿著各種情緒活動經驗，而這些經驗通常都會為決策與判斷帶來深遠的影響。過去數十年間，各領域研究的研究者們中對於情緒化決策相關議題的探索累積了豐碩的研究成果，也提出了許多結構化分類與描述現象的理論架構，但這些現象背後的機制卻仍在爭議當中。為了替這些敘述性的理論提供更明確的量測指標，並且賦予具體的神經生理基礎，我們將對於情緒化決策歷程的探討連結到增強學習理論 (reinforcement learning theory; Sutton & Barto, 1998) 的架構之中，並依循 Loewenstein 與 Lerner (2003) 的定義將情緒活動區分成伴隨性情緒 (incidental affect) 與整合性情緒 (integral affect)，分別於研究一與研究二中探討兩種情緒活動對於酬賞基礎決策 (reward-based decision making) 歷程的影響。在研究一當中，我們以中性、生氣、與高興三種不同的臉孔表情做為促發刺激嵌入機率博弈作業 (probabilistic gambling task) 當中，操弄受試者在作業過程中的伴隨性情緒活動；而在研究二當中，我們則是安排受試者在進行機率博弈作業之前先參與一個修改過後的獨裁者遊戲，並於其中操弄不同程度的分配不公平性，以引發受試者間不同的整合性情緒經驗。受試者於機率博弈作業當中的選擇行為資料以增強學習模型 (Sutton & Barto, 1998; Li, Lai, Liu, & Hsu, 2014) 配適，藉此提取與決策判斷歷程有關的模型估計參數；而實驗過程中所記錄到的腦波活動資料 (electroencephalogram, EEG) 也分析提取成事件關聯電位 (event-related potential, ERP)，以反映酬賞預測誤差 (reward-prediction error, RPE)－預期結果與實際結果之間的差距－所引發的神經活動。綜合行為、模型參數估計、以及神經生理指標量測的結果，我們發現伴隨性情緒活動與整合性情緒經驗對決策歷程的影響不盡相同，並且分別對應到不同的現行理論架構，我們將其彙整討論於最後一章。另外，文獻指出思覺失調症 (schizophrenia) 患者在酬賞處理與情緒辨識上皆有著明顯缺損，然而兩者之間的關係卻鮮少被討論，於是在研究三中我們便將研究一的設計修改後應用於這群患者上。結果顯示，在酬賞處理與情緒辨識的異常表現之餘，患者們的情緒活動對於決策力將的調節效果也明顯低於健康對照組。結合過去文獻，我們主張情緒與決策共同反應著重要的社會功能面向，而對於情緒化決策缺損的探討可增進我們對思覺失調症患者社會功能障礙的了解。

Affective experience pervades our daily life and it usually brings out pronounce impact on our judgment and decision making. However, after decades of research in the fields of social psychology, organizational psychology, and cognitive neuroscience, the mechanisms contributing to this pronounce impact is yet debating. To provide more formative indices to the existing descriptive models and to link them to more concrete neurophysiological substrates, we have examined the affective influences on decision making under the framework of reinforcement learning theory. Specifically, we followed the definition from Loewenstein and Lerner (2003) and examined the impacts of incidental and integral affects on reward-based decision making in Study 1 and Study 2, respectively. In Study 1, the incidental affect or mood state was induced through introducing different facial expressions (i.e., neutral, angry, and happy) into a probabilistic gambling task. By contrast, in Study 2, the integral affect or emotional experience was elicited through manipulating different levels of perceived unfairness in a modified dictator game preceding a probabilistic gambling task. The choice behavior data from the probabilistic gambling task in the two studies were fit with a reinforcement learning model (Sutton & Barto, 1998; Li et al., 2014) to retrieve the parameters that help us to uncover the mental processes underlying decision making. Also, in each experimental condition, the concurrent electroencephalogram (EEG) data were analyzed to retrieve the event-related potentials (ERP) that reflect the neural signaling of reward prediction error (RPE) – a discrepancy between the predicted and actual action outcomes. In addition, the schizophrenic patients are characterized with deficient reward processing and attenuated affective response, and it is of interest to investigate the altered affective influence in this population. Hence, in Study 3, we modified the design of Study 1 and applied it to chronic schizophrenic patients. This is an exploratory research and we set to investigate the potential interaction between deficient reward processing and aberrant affective activity (i.e, expression recognition) in this population, as these two factors might contribute together to the disrupted social function and poor life quality in schizophrenia. Findings of this study shall contribute to our understanding of the mechanism underlying affect-elicited decision making in schizophrenia.