Jun 2026· Engineering Interactive Computing System· 0 citations· 12 references
Computer Science
Abstract
AI-driven services continuously infer “models-of-users” to personalize content, automate decisions, and optimize interaction. Yet, most user models remain opaque, fragmented across platforms, and hard to correct, transfer, or revoke. As a result, personalisation in AI-driven services is commonly experienced as a trade-off between privacy and platform lock-in. Users must stay loyal to platforms if a model of adequate utility is to be built around them. This tension is sharpened by emerging regulatory obligations around transparency and user oversight (e.g., GDPR and the EU AI Act). We propose a full-day EICS 2026 workshop that focuses on engineering cross-compatible personalization: a vision for future interactive systems in which users can make their digital self-representations inspectable, editable, and selectively shareable. The workshop brings together researchers and practitioners from interactive systems engineering, HCI, AI/ML, and security/privacy to (1) map the design and engineering space of user-sovereign personalization; (2) derive reusable artifacts for the EICS community, such as a reference architecture, protocol sketches, and a pattern language of interaction techniques; and (3) seed a community around building interoperable toolchains for AI personalization that is not only accurate, but also controllable, auditable, and context-adaptive.
The first large-scale, cross-platform study of plugins from five major web application marketplaces, covering domains from office productivity to software development, indicates that AI-assisted plugins face a range of emerging issues that negatively impact user experience and fail to comply with established AI ethics principles.
Liuhuo Wan, Zicong Liu, Chuan Yan et al.· Proceedings of the ACM on So...· 0 citations
This proposal is the fourth edition of a workshop previously held at EICS 2023, 2024 and 2025. This edition aims to bring together researchers and practitioners interested in the engineering of interactive systems that embed AI technologies (e.g. AI-based recommender systems) or that use AI during the engineering lifecycle of interactive systems. The overall objective is to identify (based on experience reported by participants) methods, techniques, and tools to support the use and inclusion of AI technologies throughout the engineering lifecycle for interactive systems. A specific focus is on guaranteeing that user-relevant properties such as usability and user experience are accounted for. Contributions are also expected to address system-related properties, such as resilience, dependability, reliability, safety, security or performance. Another focus is on the identification and definition of software architectures supporting this integration in a generic way, ensuring properties such as maintainability and reusability.
José Creissac Campos, Camille Fayollas, Kris Luyten et al.· Engineering Interactive Comp...· 0 citations
As LLM agents evolve from single-user assistants into shared organizational infrastructure, new privacy risks emerge: inappropriate information may not only be exposed through outputs for external recipients, but also internally across users through inter-agent messages, shared memory and agents. These data spillage risks are not captured by existing privacy benchmarks grounded in contextual integrity (CI) as they focus primarily on either single-user settings or interactions between independently owned agents. We introducePiSAs (Privacy in Shared Agentic systems), a benchmark for assessing unintentional leaks with dual CI annotations: whether an information is appropriate for the task, and which users may legitimately access it. This enables direct measurement of cross-user spillage across agentic system components and interfaces, such as outputs, inter-agent communication, and memory. PiSAsis system-agnostic and supports evaluation across different agent topologies and memory regimes. We find that, although system design improves CI compliance, results are bottlenecked by incorrect LLM judgment calls: even state-of-the-art models fail to reliably filter inappropriate content or restrict transmission to authorized users. Our findings underscore the need for privacy-preserving strategies, beyond those studied in this work.
Shubham Gupta, N. Sepahvand, Abhinav Kumar et al.· 0 citations
Trustworthy deployment of GUI agents in ubiquitous computing settings requires alignment that survives dynamic interaction and precise threat conditions, not just single-turn refusal of explicit harmful requests. We argue that prompt-level alignment, the dominant lightweight defense in current mobile agents, is a local phenomenon: it works reliably only in the narrow evaluation slice where it is typically measured, namely single-turn, explicitly-verbalized intent, and degrades systematically along two axes that any real user can traverse. Using a paired diagnostic on three frontier GUI agents, screen-grounded, user-side persuasion, with no environment injection, we show that a one-line guardrail achieves large single-shot ASR reductions, up to roughly 40 points, at near-zero over-refusal cost. Nevertheless, moving from independent probes to four-turn escalation chains raises guarded ASR by approximately 20 points on every model. Relative to the neutral baselines, this increase reflects substantial guardrail erosion for Qwen but a largely defense-orthogonal dynamic risk for Claude and GPT. The sign of the salience gap flips under the guardrail: concealed requests are not systematically more successful than explicit ones without a guardrail, but are more successful with one, indicating that the defense engages primarily when intent is named. Static single-turn ASR therefore overstates deployed robustness by a systematic and predictable margin.
Haoxin An, Yunpeng Song, Zihao Bai et al.· 0 citations
The rapid advancement of LLMs has opened new opportunities in automated software engineering, driving progress in code understanding, agent-based workflows, and productivity tools. However, existing code intelligence systems have largely sidelined the end-users they aim to serve—the developers themselves. Developers exhibit substantial heterogeneity across multiple dimensions: coding style, toolchain preferences, domain-specific expertise, and problem-solving strategies. Failing to account for these individual differences directly compromises both the effectiveness of code intelligence and the likelihood of its adoption. For example, a senior architect and a junior engineer ask: "Describe the authorization module." Without personalized context, the system produces a uniform response—verbose for the expert, incomprehensible for the novice. This gap motivates a fundamental shift: from one-size-fits-all to one-size-fits-one code intelligence. A developer's dynamic in-IDE behaviors—code authoring patterns, navigation pathways, debugging trajectories—implicitly encode a rich representation of their competencies and habits. If captured and interpreted systematically, these signals can enable Personalized Code Intelligence, formalized as: [EQUATION] where P is the developer persona derived from IDE behaviors, injected alongside code context C and instruction ℐ.
Yuhong Liu, Yu Su, Zhipeng Peng et al.· SIGSOFT FSE Companion· 1 citation
Personalization algorithms determine what content users encounter on online platforms. Auditing these systems is difficult because independent auditors have only black-box access to the algorithms, while personalization depends on users'attributes, behavior, and evolving interaction histories. Existing auditing methods face a tradeoff: studies with real users capture realistic behavior but are costly and hard to control, whereas sock-puppet audits scale more easily but often rely on scripted behavior that limits realism. Beyond this, both approaches struggle to decouple user attributes from user behavior, limiting our ability to causally understand personalization. To address this gap, we introduce a framework for black-box audits of personalization algorithms using generative AI agents as behavioral engines for synthetic accounts. Each agent is instantiated with a fixed persona, grounded in demographic and political survey data, and interacts with a platform's content by reasoning about it and choosing actions. Because behavior is fixed within each persona while platform-visible signals such as age, gender, or location can be experimentally perturbed, our design enables counterfactual auditing of how platforms respond to user attributes. As a case study, we deploy 1,120 agents on X shortly after the 2024 U.S. election, spanning 14 personas and three counterfactual conditions, collecting over 200,000 content exposures. We find that X's algorithmic feed amplifies toxic, polarizing, political, and right-leaning content relative to the chronological feed, with amplification varying sharply by user ideology. Counterfactual analyses show that demographic signals affect content delivery in persona-dependent ways: pooled effects are largely null, while subgroup-level effects vary in direction and magnitude. Our work establishes GenAI-based agents as a new tool for algorithmic auditing.
Alessandro Morosini, Sarah H. Cen, Andrew Ilyas et al.· arXiv.org· 0 citations