年齡差異在處理環境選擇結果映射中特徵與結構擾動的決策策略

Abstract

身處於不斷變化的環境中，生物經常需要有效的處理訊息和執行動作以面對日常生存挑戰。為此，區分對生存極為重要的相關訊息或特徵非常關鍵。同時，準確地掌握環境中各種元素或概念之間的真實結構或關係也同等重要。在這些多面向的過程中，雜訊的存在可能會干擾智能體準確導航到其期望目標的能力。至關重要的是，有證據表明，老年人不僅在涉及不確定性的任務中感到困難，而且在維持任務表徵和保留任務結構方面也面臨著挑戰。然而，目前尚不清楚不同來源的不確定性如何解釋老年人在日常生活中的認知表現。在這項研究中，我們探討了神經策略應對這兩個訊息處理的不確定性時與年齡相關的潛在差異。我們首先設計了五種基於強化學習演算法的計算模型，這些模型為評估年輕人和老年人在動態不確定情境下訊息處理的差異提供了理論基礎。接下來，我們執行了一個兩階段基於謎題的功能性磁振造影（fMRI）實驗，以觀察真實人類在應對不同程度的特徵和結構不確定性時的行為和神經機制。在這項任務中，參與者需要從包含不同的顏色和形狀組合的鑰匙中選擇正確的鑰匙，以解鎖各個房間的門，從而最大化獎勵。27名年輕人（年齡 = 22.28 ± 1.69，17名女性，最小 = 20.21歲，最大 = 25.86歲）和22名老年人（年齡 = 68.63 ± 2.40，14名女性，最小 = 65.49歲，最大 = 74.03歲）的行為結果顯示，與年輕對照組相比，隨著特徵雜訊的增加，老年人的表現顯著下降，顯示其在應對特徵雜訊方面表現出次優的能力。使用計算模型對潛在認知機制的進一步分析表明，年輕人更傾向於採用無模型的演算法，而老年人則傾向於更基於模型的策略，儘管這兩個年齡組都擁有識別打開房門的相關維度的模組。最後，27名年輕人（年齡 = 22.28 ± 1.69，17名女性，最小 = 20.21歲，最大 = 25.86歲）和20名老年人（年齡 = 68.46 ± 2.27，13名女性，最小 = 65.49歲，最大 = 74.03歲）的腦影像結果顯示，結構雜訊增加會引起預設模式區域更高的反應，表示在老年人的內側額葉活化增加，以及年輕人的楔前葉活化增加。相反，特徵雜訊升高會促使非預設模式區域產生更高的反應，在老年人中，小腦和額葉的反應增加，而在年輕人中，海馬旁和顳葉中迴的反應增加。研究結果反映了年齡相關導致應對環境不確定性時不同行為的神經計算策略差異。

In ever-changing environments, organisms face daily survival challenges that demand effective information processing and action execution. To this end, it is essential to discern pertinent information or features crucial for survival. Simultaneously, accurately grasping the true structure or relationships between various elements or concepts within the environment is equally important. Amidst these multifaceted processes, the presence of noise can disrupt an agent’s ability to navigate accurately toward its desired objectives. Critically, evidence shows that older adults not only struggle with tasks involving uncertainty but also face challenges in maintaining task representation and preserving task structure. However, it remains unclear how various sources of uncertainty contribute in explaining older adult cognitive performances in daily life. In this study, we explore potential age-related differences in neural strategic coping with uncertainty in both of these two information processing facets. We first devised five computational models based on reinforcement learning algorithms. The models provided theoretical bases for assessing younger and older adult information processing differences in dynamic uncertain scenarios. Next, we conducted a two-stage puzzle-based functional magnetic resonance imaging (fMRI) experiment to observe real human behavior and neural mechanisms in response to varying levels of representational and structural uncertainty. In this task, participants selected the correct keys, each comprising different combinations of colors and shapes, to unlock doors to various rooms and thereby maximize rewards. The behavioral results from 27 younger adults (age = 22.28 ± 1.69, 17 female, minimum = 20.21 years old, maximum = 25.86 years old) and 22 older adults (age = 68.63 ± 2.40, 14 female, minimum = 65.49 years old, maximum = 74.03 years old) revealed that older adults exhibit a suboptimal coping with noise in features compared to their younger counterparts, as evidenced by a significant performance decline as noise in features increased. A further analysis of the underlying cognitive mechanisms, using the computational models, suggested that younger adults favored a more model-free algorithm, while older adults tended toward a more model-based strategy, despite both age groups possessing the module to identify the relevant dimensions of keys to open room doors. Finally, brain imaging results from 27 younger adults (age = 22.28 ± 1.69, 17 female, minimum = 20.21 years old, maximum = 25.86 years old) and 20 older adults (age = 68.46 ± 2.27, 13 female, minimum = 65.49 years old, maximum = 74.03 years old) revealed that increased noise in structure induced higher responses in default mode areas, with heightened medial frontal activation in older adults but increased precuneus activation in younger adults. Conversely, elevated noise in feature prompted higher responses in non-default mode areas, with increased cerebellar and frontal responses in older adults and heightened parahippocampal and middle temporal responses in younger adults. The findings reflect age-related differences in neurocomputational strategies that yield distinct behaviors in response to environmental uncertainty.