Simret Araya Gebreegziabher

Simret Araya Gebreegziabher, Allison E Sproul, Yinuo Yang et al.

Current human-AI alignment and evaluation methods for large language models (LLMs) often rely on preference signals collected immediately after an interaction. This practice implicitly treats preference as static, even though many LLM-mediated decisions unfold over time and may be re-evaluated differently after real-world consequences and observed outcomes. Therefore, we argue for a methodological shift from single-moment preference elicitation to longitudinal, context-situated alignment measurement. We present a methodological framework for collecting temporally grounded alignment signals by combining (1) in-situ preference capture, (2) context-triggered follow-up preference reflection, and (3) privacy-preserving behavioral traces that help interpret preference change. As an instantiation of this methodology, we introduce BITE, a browser-based system that detects consequential LLM interactions, prompts reflection across later decision points, and supports progressive, user-controlled consent for sharing behavioral data. Through a two week longitudinal deployment study with 8 participants, our approach surfaced differences between immediate and later user preferences in accuracy, relevance and other dimensions of the LLM output. Our findings highlight the limitations of single-moment preference datasets and underscore the importance of longitudinal methods for alignment evaluation in everyday use.

Chaoran Chen, Zhiping Zhang, Zeya Chen et al.

LLM-powered computer-use agents (CUAs) are shifting users from direct manipulation to supervisory coordination. Existing oversight mechanisms, however, have largely been studied as isolated interface features, making broader oversight strategies difficult to compare. We conceptualize CUA oversight as a structural coordination problem defined by delegation structure and engagement level, and use this lens to compare four oversight strategies in a mixed-methods study with 48 participants in a live web environment. Our results show that oversight strategy more reliably shaped users' exposure to problematic actions than their ability to correct them once visible. Plan-based strategies were associated with lower rates of agent problematic-action occurrence, but not equally strong gains in runtime intervention success once such actions became visible. On subjective measures, no single strategy was uniformly best, and the clearest context-sensitive differences appeared in trust. Qualitative findings further suggest that intervention depended not only on what controls users retained, but on whether risky moments became legible as requiring judgment during execution. These findings suggest that effective CUA oversight is not achieved by maximizing human involvement alone. Instead, it depends on how supervision is structured to surface decision-critical moments and support their recognition in time for meaningful intervention.

Simret Araya Gebreegziabher, Kuangshi Ai, Zheng Zhang et al.

Active Learning (AL) allows models to learn interactively from user feedback. This paper introduces a counterfactual data augmentation approach to AL, particularly addressing the selection of datapoints for user querying, a pivotal concern in enhancing data efficiency. Our approach is inspired by Variation Theory, a theory of human concept learning that emphasizes the essential features of a concept by focusing on what stays the same and what changes. Instead of just querying with existing datapoints, our approach synthesizes artificial datapoints that highlight potential key similarities and differences among labels using a neuro-symbolic pipeline combining large language models (LLMs) and rule-based models. Through an experiment in the example domain of text classification, we show that our approach achieves significantly higher performance when there are fewer annotated data. As the annotated training data gets larger the impact of the generated data starts to diminish showing its capability to address the cold start problem in AL. This research sheds light on integrating theories of human learning into the optimization of AL.

Chaoran Chen, Zhiping Zhang, Bingcan Guo et al.

A Large Language Model (LLM) powered GUI agent is a specialized autonomous system that performs tasks on the user's behalf according to high-level instructions. It does so by perceiving and interpreting the graphical user interfaces (GUIs) of relevant apps, often visually, inferring necessary sequences of actions, and then interacting with GUIs by executing the actions such as clicking, typing, and tapping. To complete real-world tasks, such as filling forms or booking services, GUI agents often need to process and act on sensitive user data. However, this autonomy introduces new privacy and security risks. Adversaries can inject malicious content into the GUIs that alters agent behaviors or induces unintended disclosures of private information. These attacks often exploit the discrepancy between visual saliency for agents and human users, or the agent's limited ability to detect violations of contextual integrity in task automation. In this paper, we characterized six types of such attacks, and conducted an experimental study to test these attacks with six state-of-the-art GUI agents, 234 adversarial webpages, and 39 human participants. Our findings suggest that GUI agents are highly vulnerable, particularly to contextually embedded threats. Moreover, human users are also susceptible to many of these attacks, indicating that simple human oversight may not reliably prevent failures. This misalignment highlights the need for privacy-aware agent design. We propose practical defense strategies to inform the development of safer and more reliable GUI agents.

Chaoran Chen, Zhiping Zhang, Ibrahim Khalilov et al.

The rise of Large Language Models (LLMs) has revolutionized Graphical User Interface (GUI) automation through LLM-powered GUI agents, yet their ability to process sensitive data with limited human oversight raises significant privacy and security risks. This position paper identifies three key risks of GUI agents and examines how they differ from traditional GUI automation and general autonomous agents. Despite these risks, existing evaluations focus primarily on performance, leaving privacy and security assessments largely unexplored. We review current evaluation metrics for both GUI and general LLM agents and outline five key challenges in integrating human evaluators for GUI agent assessments. To address these gaps, we advocate for a human-centered evaluation framework that incorporates risk assessments, enhances user awareness through in-context consent, and embeds privacy and security considerations into GUI agent design and evaluation.

Ruyuan Wan, Simret Gebreegziabhe, Toby Jia-Jun Li et al.

In recent years, there has been a growing interest in employing intelligent agents in writing. Previous work emphasizes the evaluation of the quality of end product-whether it was coherent and polished, overlooking the journey that led to the product, which is an invaluable dimension of the creative process. To understand how to recognize human efforts in co-writing with intelligent writing systems, we adapt Flower and Hayes' cognitive process theory of writing and propose CoCo Matrix, a two-dimensional taxonomy of entropy and information gain, to depict the new human-agent co-writing model. We define four quadrants and situate thirty-four published systems within the taxonomy. Our research found that low entropy and high information gain systems are under-explored, yet offer promising future directions in writing tasks that benefit from the agent's divergent planning and the human's focused translation. CoCo Matrix, not only categorizes different writing systems but also deepens our understanding of the cognitive processes in human-agent co-writing. By analyzing minimal changes in the writing process, CoCo Matrix serves as a proxy for the writer's mental model, allowing writers to reflect on their contributions. This reflection is facilitated through the measured metrics of information gain and entropy, which provide insights irrespective of the writing system used.

Jingyu Tang, Chaoran Chen, Jiawen Li et al.

The dark patterns, deceptive interface designs manipulating user behaviors, have been extensively studied for their effects on human decision-making and autonomy. Yet, with the rising prominence of LLM-powered GUI agents that automate tasks from high-level intents, understanding how dark patterns affect agents is increasingly important. We present a two-phase empirical study examining how agents, human participants, and human-AI teams respond to 16 types of dark patterns across diverse scenarios. Phase 1 highlights that agents often fail to recognize dark patterns, and even when aware, prioritize task completion over protective action. Phase 2 revealed divergent failure modes: humans succumb due to cognitive shortcuts and habitual compliance, while agents falter from procedural blind spots. Human oversight improved avoidance but introduced costs such as attentional tunneling and cognitive load. Our findings show neither humans nor agents are uniformly resilient, and collaboration introduces new vulnerabilities, suggesting design needs for transparency, adjustable autonomy, and oversight.