讲座题目：Preference Elicitation for Algorithmic Personalization

主 讲 人：盖嘉 北京大学光华管理学院助理教授

讲座时间：2025年5月14日10：00

讲座地点：学院南路校区主教学楼 615

主 持 人：魏闯 中央财经大学商学院助理教授

主讲人介绍：

盖嘉，北京大学光华管理学院市场营销系助理教授。研究领域为消费者行为，包括数字消费、道德行为、与自控力。研究成果发表于Journal of Marketing, Journal of Consumer Research, Journal of Experimental Psychology: General等期刊。加入光华前，先后获得香港中文大学心理学学士，芝加哥大学社会科学硕士，以及荷兰鹿特丹商学院市场（伊拉斯姆斯大学）营销学博士学位。

讲座摘要：

Companies and platforms frequently request consumers to indicate their preferences to generate personalized content, assuming that users will reveal their full range of interests. However, a series of experiments, including several conducted on custom-built platforms, reveals that consumers often omit interests that they would otherwise select or consider when sharing their preferences with algorithms. This tendency arises because 1) consumers believe that sharing narrow preferences reduces the risk of being misclassified by algorithms, and 2) they do not adequately consider all predefined categories of interest. Two separate research projects document these distinct processes. Crucially, the preferences shared by consumers create a feedback loop: those who select narrow preferences receive less diverse recommendations, which further narrows their choices and subsequent recommendations. Our findings thus delineate the conditions under which filter bubbles are generated and dismantled, suggesting that simple yet effective design features can significantly alter the content that individuals are exposed to and engage with.

摘要（中文）：

算法依赖于消费者的偏好来生成个性化推荐。其背后的假设是消费者数据代表了他们真正的兴趣。因此，“信息茧房”、“回音室”等问题常被归因于消费者对多样内容的抗拒。本研究挑战了这一观点。我将通过两个研究项目展示：算法推荐的偏好测量可能导致人们选择性展示狭窄的、特定的兴趣。在应用场景中，这些偏好被算法进一步放大，导致截然不同的观看内容。