檢驗跨時與冒險選擇之雙曲線乘積典範在人類志願者的適用性

Testing the Applicability of a Multiplicative Hyperbolic Model of Inter-Temporal and Risky Choice in Human Volunteers

背景：衝動行為可被定義為偏好選擇小且立即能獲得的酬賞，而捨棄大但有延宕的酬賞；或一種偏好冒險的傾向。以雙曲線乘積典範（Ho, Mobini, Chiang, Bradshaw, & Szabadi, 1999）之量化方法分析跨時選擇的結果證實，此典範可成功地描述動物的衝動或冒險行為。本研究的目的在檢驗此典範是否可適用於人類。方法：四十名健康志願者均參與兩個由兩種選項（A或B）中選擇金錢酬賞的實驗作業。兩項作業採相同的設計，分別包含數個實驗情境。在一作業下，選項A的酬賞會依特定的機率隨機出現，但酬賞出現的延宕時間（dA）較短，而選項B的酬賞則是一定會出現，但出現的延宕時間（dB）卻較長。每個作業共分成5個分別包含5種不同dA值的情境。每個情境各有50個嘗試（10 X 5嘗試種類）。根據每位參與者於每一個情境中的反應可分別估計出其在該情境的主觀相等點dB（50））（即參與者於該情境中選擇B的反應等於50%時所對應的延宕時間）。在另一作業下，選擇A或B後所得到的酬賞大小不同，但兩種酬賞出現的機率則是相等的，其餘的實驗程序均與上述的作業類似。本研究利用線性迴歸的方法分別檢驗兩項作業中與dA與dB（50）之間的關係，並以這些線性關係中的斜率與截距分別代表個體對增強大小、延宕增強、與機率增強的敏感度。結果：在兩項作業中，dB（50）均以線性的趨勢隨著dA的增加而逐步遞增（r2 > .99）。兩作業中所得出之線性函數的截距（延宕增強之敏感度）有顯著的正相關（r = .60, p < .001），但兩線性函數之間的斜率（分別代表機率增強與增強大小的敏感度）的相關則不顯著（r = .18,p = .269）。結論：本研究的結果顯示，雙曲線乘積典範中用以說明酬賞物大小、延宕、與出現機率對選擇之效用的線性等式似乎亦可於描述人類的資料。

Background: Impulsive choice can be defined as the selection of small immediate rewards rather than larger delayed rewards, or a predisposition to risk taking. A multiplicative hyperbolic model (Ho, Mobini, Chiang, Bradshaw, & Szabadi, 1999) proposes a quantitative methodology for analyzing inter-temporal choice has proved successful in describing impulsive and risky choice in rats. The present study aims to test the applicability of the model to data from human participants. Methods: Forty healthy volunteers underwent two experimental tasks of choosing between two alternatives (A and B)for monetary rewards. Each task consisted of several conditions based on the same design. On one task, alternative A produced a probabilistic reward after a short delay (dA)，and B a certain reward after longer delays, dB, dA was manipulated across 5 conditions. There were 50 trials (10 x 5 trial type) in each condition. Indifference delays, dB(50) (value of dB yielding 50% choice of B)were estimated for each participant in each condition. On the other task, a similar procedure was employed, except that reward sizes for A and B were different, but with the same probabilities. Linear functions of dB(50) versus dA were fitted; the slopes and intercepts provided indices of sensitivity to reinforcement size, delay and probability. Results: dB(50) increased linearly with dA (r2 > .99) in the two tasks. There was a significant correlation (r = .60, p < .001) between the intercepts (sensitivity to delay) but not between the slopes (sensitivity to size and probability) in the two tasks (r = .18, p = .269). Conclusion: These results show the indifference equations specified by the multiplicative model to account for effects of reward size, delay and probability on choice appear to be applicable to data from humans.