2023-08-012024-05-13https://scholars.lib.ntu.edu.tw/handle/123456789/652627本計畫從「描述」、「解釋」、「預測」、「控制」這四個面向探討當今熱門的幸福研究之根本性難題，並針對這些難題一一提出可行的解決方法。在「描述」方面 的難題是稀疏且沒有效度的量測造成不符實情的圖像，而基於生理心理學的解方是 改用密集且有效度的量測(如心跳率與心跳變異率)。在「控制」方面的難題是無法 預知特定介入在一個體上的有效性，而基於控制理論的解方是透過最佳化的實驗 (如Bandit演算法)找出對個體有效的介入。在「預測」方面的難題是無法捕捉各種 混淆變項造成預測壓力動態不準，而基於動態系統理論的解方是利用個體動態特性 來做預測(如利用Takens’ Embedding Theorem)。在「解釋」方面的難題是簡化的 線性模型可能給出不正確的解釋，而基於統計學習理論的解方是改用非線性的機器學習模型來幫助推論因果(如利用可解釋性人工智慧或是因果機器學習方法)。本計畫共規劃四年的時間來執行八個子計畫，而最終的產品將會是透過智慧型手機或是機器人，能提供個人化的有效減壓策略，並能讓此策略受到個體採納的推薦系統。 This proposal addresses the fundamental problems in today`s well-being research from the aspects of "description", "explanation", "prediction", and "control". It proposes feasible solutions to each of these problems. The difficulty in "description" is that sparse and invalid measurements lead to a distorted picture of reality, and the solution based on biological psychology is to carry out dense and valid measurements (e.g., heart rate and heart rate variability). The difficulty in "control" is that it is impossible to predict the effectiveness of a specific intervention on an individual, and the solution based on control theory is to find an effective intervention for an individual through optimized experiments (e.g., via Bandit algorithm). The difficulty in "prediction" is that it is impossible to capture various confounding variables that lead to inaccurate predictions, and the solution based on dynamic system theory is to use individual dynamic characteristics to make predictions (e.g., leveraging Takens` Embedding Theorem). The difficulty in "explanation" is that simplified linear models may give incorrect explanations, and the solution based on statistical learning theory is to use non-linear machine learning models to help infer causality (e.g., using methods of explainable artificial intelligence or causal machine learning). This proposal plans a total of four years to carry out eight sub-projects, and the end product will be a recommendation system embedded in smartphones or social robots that can provide not only personalized but also effective stress reduction strategies adopted by individuals.幸福;個體差異;壓力;心跳率;心跳變異率;介入;控制理論;人工智慧;機器學習;深度學習;推薦系統;因果推論;Well-being; Individual Difference; Stress; Heart Rate; Heart Rate Variability; Intervention; Artificial Intelligence; Machine Learning; Deep Learning; Recommendation System; Causal Inference