CHS: Medium: Immersive Recommendation Systems: User-Centric Recommendation Models and Applications

CHS：中：沉浸式推荐系统：以用户为中心的推荐模型和应用

• 批准号: 1700832
• 负责人: Deborah Estrin
• 金额: $ 120万
• 依托单位: Cornell University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-08-01 至 2022-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1700832&HistoricalAwards=false
• 关键词: CHS; Medium; Immersive; Recommendation; Systems

This work strives to create a future in which hyper-personalized content, digital services, and personal information management tools let individuals benefit from the data they generate more directly, selectively, and transparently. Individuals will then be empowered to gain insights into their own behavior, personalize their own experiences, and ultimately more effectively utilize the services to achieve their goals. Moreover, the systems that engage end-users with the data they generate can promote local processing and selective sharing of personal information. Given the pervasiveness of online tools in a person's work, personal, and social life, the actions they take are increasingly shaped by recommendation systems. Today's approaches to personalization and recommendation are provider centric. Society as a whole will benefit from broader exploration of personalization and recommendation from the consumers' perspective. Also along these lines, there is increasing concern about the shifts in expectations for individual control over data sharing. This project's user-centered and personal-sharing-policy-aware design is a potential solution to address this tension between providers and users, allowing people to more directly benefit from their data.The research objective is to develop novel user modeling techniques, policy-aware systems, and rich user interactions that allow individuals to harness their own diverse digital traces ("small data"), enable novel applications, and receive more personally-relevant recommendations while limiting privacy exposure. This research will contribute the novel user modeling and interaction techniques needed to put the individual at the center of their personalization, in particular: (1) Immersive user modeling techniques that analyze diverse types of user data, including social media streams, private text communications, web browsing, geo-location traces, and personal images, to incorporate users' diverse and idiosyncratic interests. (2) Novel recommendation models and policy-aware software architecture that consists of open source building blocks designed to facilitate generalization of this approach to ingest diverse personal data traces and feed diverse application targets. (3) Methods to understand and address key human-centered challenges in immersive recommender systems through participatory design of the user experience, as well as qualitative and quantitative evaluation of deployed systems and applications.

这项工作致力于创造一个未来，在这个未来中，超个性化的内容、数字服务和个人信息管理工具可以让个人更直接、更有选择性和更透明地从他们生成的数据中受益。这样，个人就可以洞察自己的行为，个性化自己的体验，并最终更有效地利用这些服务来实现自己的目标。此外，让终端用户参与其生成的数据的系统可以促进本地处理和选择性地共享个人信息。鉴于在线工具在人们的工作、个人和社会生活中无处不在，他们采取的行动越来越多地受到推荐系统的影响。今天的个性化和推荐方法是以提供者为中心的。从消费者的角度对个性化和推荐进行更广泛的探索，将使整个社会受益。与此同时，人们越来越担心个人对数据共享控制权的期望会发生变化。该项目以用户为中心和个人共享策略感知的设计是解决提供者和用户之间紧张关系的潜在解决方案，允许人们更直接地从他们的数据中受益。研究目标是开发新颖的用户建模技术、策略感知系统和丰富的用户交互，使个人能够利用自己的各种数字痕迹（“小数据”），启用新颖的应用程序，并在限制隐私暴露的同时接收更多与个人相关的建议。本研究将提供新颖的用户建模和交互技术，将个人置于个性化的中心，特别是：(1)沉浸式用户建模技术，分析不同类型的用户数据，包括社交媒体流、私人文本通信、网页浏览、地理位置跟踪和个人图像，以结合用户的多样化和特殊兴趣。(2)新颖的推荐模型和策略感知软件架构，该架构由开源构建块组成，旨在促进该方法的推广，以摄取不同的个人数据跟踪并提供不同的应用目标。(3)通过用户体验的参与式设计，以及对部署的系统和应用进行定性和定量评估，了解和解决沉浸式推荐系统中以人为中心的关键挑战的方法。

项目成果

RecBoard: A Web-based Platform for Recommendation System Research and Development

• DOI: 10.1145/3308558.3314133
• 发表时间: 2019-05
• 期刊: The World Wide Web Conference
• 作者: M. Chawla;Kriti Singh;Longqi Yang;D. Estrin

How Intention Informed Recommendations Modulate Choices: A Field Study of Spoken Word Content

意图告知的建议如何调节选择：口语内容的实地研究

• DOI: 10.1145/3308558.3313540
• 发表时间: 2019
• 期刊: World Wide Web Conference (The Web Conference
• 作者: Yang, Longqi;Sobolev, Michael;Wang, Yu;Chen, Jenny;Dunne, Drew;Tsangouri, Christina;Dell, Nicola;Naaman, Mor;Estrin, Deborah
• 通讯作者: Estrin, Deborah

Spinning Language Models: Risks of Propaganda-As-A-Service and Countermeasures

• DOI: 10.1109/sp46214.2022.9833572
• 发表时间: 2021-12
• 期刊: 2022 IEEE Symposium on Security and Privacy (SP)
• 作者: Eugene Bagdasaryan;Vitaly Shmatikov

Unbiased offline recommender evaluation for missing-not-at-random implicit feedback

• DOI: 10.1145/3240323.3240355
• 发表时间: 2018-09
• 期刊: Proceedings of the 12th ACM Conference on Recommender Systems
• 作者: Longqi Yang;Yin Cui;Yuan Xuan;Chenyang Wang;Serge J. Belongie;D. Estrin

Ancile: Enhancing Privacy for Ubiquitous Computing with Use-Based Privacy

• DOI: 10.1145/3338498.3358642
• 发表时间: 2019-11
• 期刊: Proceedings of the 18th ACM Workshop on Privacy in the Electronic Society
• 作者: Eugene Bagdasaryan;Griffin Berlstein;J. Waterman;Eleanor Birrell;Nate Foster;F. Schneider;D. Estrin