內感覺、網路使用與情緒調節在決策作業預期性膚電反 應之整合性影響

Abstract

本論文旨在分析內感覺能力、情緒調節方式以及網路使用行為等個體差異，如何共同影響決策過程中所出現的預期性生理反應。過去文獻指出，個體的身體覺察能力、網路使用行為與情緒調節策略，皆可能經由調節預期性生理訊號，進而影響決策表現。其中，膚電反應（skin conductance responses, SCRs）被廣泛作為預期性生理訊號的操作性指標，理論上被視為軀體標記（somatic markers）之運作機制。上述三項個體差異變項雖在實證研究中已展現一定程度的交互關聯，但目前尚缺乏針對這些因素如何透過預期性膚電反應的產生與行為應用，對決策歷程產生整合性作用的系統性探討。本論文以124名受試者為研究對象，整合性評估內感覺（採用多向度內感覺覺察量表）、內感覺準確性、覺察與信心水準（以心跳計數作業為測量工具）、網路使用行為（包含網路成癮量表及自陳每週上網時數）以及情緒調節傾向（情緒調節問卷），探討這些變項如何共同影響愛荷華賭局作業（Iowa Gambling Task, IGT）中的預期性膚電反應（aSCRs），並進一步檢驗aSCRs與IGT表現間之關聯。本文針對兩類決策情境進行分析與比較：其一為期望值／風險情境，選項之期望值呈系統性差異（不利牌組A+B相對有利牌組C+D）；其二為損失頻率情境，在平均期望值相等的條件下比較損失頻率結構（低頻率損失牌組B+D相對高頻率損失牌組A+C），並涵蓋IGT三個階段的歷程。結果顯示，個體差異預測變項與aSCRs之間的關聯會因決策情境的不同而有所變化，並且這種關係會隨著IGT階段的進展而有所差異。透過將個體差異預測變項進行 K-平均數集群分析，可區辨出參與者具有三種異質的個體差異輪廓，並發現其aSCRs展現出顯著且有系統的差異。在三者之中，集群A有最高的內感覺覺察與準確性，情緒調節方面更傾向運用認知再評估策略，且網路成癮傾向較低；集群Ｂ內感覺覺察與準確性中等、認知再評估傾向介於集群A與C之間、網路使用行為與成癮傾向較集群A高；集群C內感覺覺察與信心較低、較不使用認知再評估策略，且網路使用行為與成癮傾向較高。具體而言，參與者針對低頻率損失牌組（B+D）的aSCRs於三個IGT階段皆展現出穩定的集群排序；集群Ａ對於B+D牌組的aSCR反應最為顯著，而集群C則反應最弱。反之，頻率差值指標（B+D − A+C）與高頻率損失牌組（A+C）之aSCRs皆呈現出明顯的集群與階段交互作用，集群之間的差距會隨著IGT階段擴大、縮小，或在部分階段出現重新排列。這個發現顯示與頻率相關aSCR的形成與隨IGT階段的調整，會因個體差異輪廓而呈現不同軌跡。同樣地，在期望值／風險決策情境中，不利牌組（A+B）與有利牌組（C+D）及其差值指標（A+B − C+D）之aSCRs同樣隨著IGT階段展現集群間的顯著變化；特別值得注意的是，集群C於IGT中期在期望值／風險差值指標上出現方向性反轉，表示其aSCR不再呈現典型的「對不利牌組（A+B）較強、對有利牌組（C+D）較弱」的風險預期模式；相反地，其反應在中期更偏向對有利牌組或非典型線索產生較高的預期性喚起，與其他集群存在顯著差異。在行為層面，期望值／風險決策情境下的aSCR差值與較高比例的有利選擇行為呈現顯著相關；然而，在損失頻率情境中，aSCR與選擇行為之間的關聯性則根據不同集群而有所差異，顯示不同個體差異的參與者在生理訊號與決策行為之間展現不同的耦合模式。本研究結果以情境化且以個體差異為核心的分析策略支持損失頻率相關與期望值／風險相關的預期性生理訊號並非可互換的單一指標，提供對建立決策歷程模型及探討決策之預期性機制的參考。

This thesis examines how individual differences in interoceptive processing, emotion regulation, and internet-use tendencies jointly shape anticipatory physiological signaling during decision-making under uncertainty. Prior research has demonstrated that individuals’ bodily awareness, internet usage patterns, and emotional regulation strategies are connected to decision-making by affecting anticipatory physiological signals, operationalized via anticipatory skin conductance responses (aSCRs), theorized to function as somatic markers. These three individual differences are also known to be interrelated. However, no study has investigated the combined effects of these factors on decision-making through the creation and use of aSCRs. This thesis addresses this gap using data from N = 124 participants. Measures of interoceptive sensibility, accuracy, awareness, and confidence, internet usage patterns, and emotion regulation tendencies were used to derive participant profiles via k-means clustering. We then examined aSCRs in the Iowa Gambling Task (IGT) across three stages in two decision contexts: expected value/risk (advantageous C+D vs disadvantageous A+B) and loss-frequency (low-frequency loss decks B+D vs high-frequency loss decks A+C). Results revealed dissociation between loss-frequency and expected-value aSCRs. K-means clustering identified three heterogeneous profiles: Cluster A showed the highest interoceptive sensibility and accuracy, greater reliance on cognitive reappraisal, and lower problematic internet use; Cluster B showed intermediate interoceptive and regulatory resources alongside higher internet engagement than Cluster A; Cluster C showed lower interoceptive sensibility and confidence, less reappraisal use, and comparatively elevated problematic internet use. In the loss-frequency context, profiles differed reliably in aSCR magnitude to low-frequency loss decks (B+D), yielding a stable ordering across stages (Cluster A > Cluster B > Cluster C). In contrast, aSCRs to high-frequency loss decks (A+C) and the frequency difference (B+D −A+C) showed a clear Cluster × Stage interaction where the size and pattern of between-profile differences changed across stages. Similarly, in the expected value/risk context, aSCRs to disadvantageous decks (A+B), advantageous decks (C+D), and the EV contrast (C+D − A+B) varied across stages in a profile-dependent manner. Behaviorally, risk-related aSCR differentiation was associated with more advantageous choices, whereas in the loss-frequency context relationships differed by profile, suggesting heterogeneous coupling between physiology and behavior. Dimensions of interoception served distinct functions: perceptual-attentional facets predicted overall anticipatory amplitude, embodied trust facets predicted context-appropriate allocation. Overall, findings support a context-specific, individual-difference-centered account in which loss-frequency and risk-related signals are not interchangeable and are shaped by partially distinct learning-dependent dynamics.