Search for: All records

Visual content shared on social media often includes sensitive elements that can threaten personal privacy. While privacy protection tools--some of which are powered by the state-of-the-art generative AI (Gen-AI) technologies--have been increasingly developed to address such visual privacy concerns by identifying sensitive elements in visual content and suggesting or applying modifications to process the visual content, the success of these tools depends on how well they meet users' nuanced needs and preferences. In this study, we conducted semi-structured interviews with 18 individuals who have either experienced or caused privacy violations in shared visual content in the past to gather first-hand perspectives on stakeholders' privacy concerns, their preferences for how to address these concerns, and their attitude toward the use of generative AI for privacy protection. Our findings highlight that sensitive elements are often not limited to direct identifiers but include contextual combinations and external information that can lead to unintended inferences. Decisions about whether and what to modify are shaped by concerns about privacy effectiveness, content value, content meaning, and emotional or social relevance, while choices around how to modify are influenced by recognition difficulty, visual content integrity, contextual consistency, atmosphere, and usability of modification methods. Participants saw Gen-AI as a promising tool for lowering editing barriers and enhancing creative control but also raised concerns about data usage, manipulation, and transparency. Importantly, we identify tensions between uploaders and depicted individuals, emphasizing the need for shared consent mechanisms and user-centered design in privacy protection. We conclude by discussing design implications for context-aware, flexible, and ethically responsible privacy tools. 

AI-assisted decision making becomes increasingly prevalent, yet individuals often fail to utilize AI-based decision aids appropriately especially when the AI explanations are absent, potentially as they do not reflect on AI’s decision recommendations critically. Large language models (LLMs), with their exceptional conversational and analytical capabilities, present great opportunities to enhance AI-assisted decision making in the absence of AI explanations by providing natural-language-based analysis of AI’s decision recommendation, e.g., how each feature of a decision making task might contribute to the AI recommendation. In this paper, via a randomized experiment, we first show that presenting LLM-powered analysis of each task feature, either sequentially or concurrently, does not significantly improve people’s AI-assisted decision performance. To enable decision makers to better leverage LLM-powered analysis, we then propose an algorithmic framework to characterize the effects of LLM-powered analysis on human decisions and dynamically decide which analysis to present. Our evaluation with human subjects shows that this approach effectively improves decision makers’ appropriate reliance on AI in AI-assisted decision making. 

Recent advances in AI models have increased the integration of AI-based decision aids into the human decision making process. To fully unlock the potential of AI-assisted decision making, researchers have computationally modeled how humans incorporate AI recommendations into their final decisions, and utilized these models to improve human-AI team performance. Meanwhile, due to the black-box'' nature of AI models, providing AI explanations to human decision makers to help them rely on AI recommendations more appropriately has become a common practice. In this paper, we explore whether we can quantitatively model how humans integrate both AI recommendations and explanations into their decision process, and whether this quantitative understanding of human behavior from the learned model can be utilized to manipulate AI explanations, thereby nudging individuals towards making targeted decisions. Our extensive human experiments across various tasks demonstrate that human behavior can be easily influenced by these manipulated explanations towards targeted outcomes, regardless of the intent being adversarial or benign. Furthermore, individuals often fail to detect any anomalies in these explanations, despite their decisions being affected by them. 

Recent advances in AI models have increased the integration of AI-based decision aids into the human decision making process. To fully unlock the potential of AI- assisted decision making, researchers have computationally modeled how humans incorporate AI recommendations into their final decisions, and utilized these models to improve human-AI team performance. Meanwhile, due to the “black-box” nature of AI models, providing AI explanations to human decision makers to help them rely on AI recommendations more appropriately has become a common practice. In this paper, we explore whether we can quantitatively model how humans integrate both AI recommendations and explanations into their decision process, and whether this quantitative understanding of human behavior from the learned model can be utilized to manipulate AI explanations, thereby nudging individuals towards making targeted decisions. Our extensive human experiments across various tasks demonstrate that human behavior can be easily influenced by these manipulated explanations towards targeted outcomes, regardless of the intent being adversarial or benign. Furthermore, individuals often fail to detect any anomalies in these explanations, despite their decisions being affected by them. 

In an era marked by rampant online misinformation, artificial intelligence (AI) technologies have emerged as tools to combat this issue. This paper examines the effects of AI-based credibility indicators in people’s online information processing under the social influence from both peers and “experts”. Via three pre-registered, randomized experiments, we confirm the effectiveness of accurate AI-based credibility indicators to enhance people’s capability in judging information veracity and reduce their propensity to share false information, even under the influence from both laypeople peers and experts. Notably, these effects remain consistent regardless of whether experts’ expertise is verified, with particularly significant impacts when AI predictions disagree with experts. However, the competence of AI moderates the effects, as incorrect predictions can mislead people. Furthermore, exploratory analyses suggest that under our experimental settings, the impact of the AI-based credibility indicator is larger than that of the expert’s. Additionally, AI’s influence on people is partially mediated through peer influence, although people automatically discount the opinions of their laypeople peers when seeing an agreement between AI and peers’ opinions. We conclude by discussing the implications of utilizing AI to combat misinformation. 

AI-assisted decision-making systems hold immense potential to enhance human judgment, but their effectiveness is often hindered by a lack of understanding of the diverse ways in which humans take AI recommendations. Current research frequently relies on simplified, ``one-size-fits-all'' models to characterize an average human decision-maker, thus failing to capture the heterogeneity of people's decision-making behavior when incorporating AI assistance. To address this, we propose Mix and Match (M&M), a novel computational framework that explicitly models the diversity of human decision-makers and their unique patterns of relying on AI assistance. M&M represents the population of decision-makers as a mixture of distinct decision-making processes, with each process corresponding to a specific type of decision-maker. This approach enables us to infer latent behavioral patterns from limited data of human decisions under AI assistance, offering valuable insights into the cognitive processes underlying human-AI collaboration. Using real-world behavioral data, our empirical evaluation demonstrates that M&M consistently outperforms baseline methods in predicting human decision behavior. Furthermore, through a detailed analysis of the decision-maker types identified in our framework, we provide quantitative insights into nuanced patterns of how different individuals adopt AI recommendations. These findings offer implications for designing personalized and effective AI systems based on the diverse landscape of human behavior patterns in AI-assisted decision-making across various domains. 

Mobile devices such as smartphones, laptops, and tablets can often connect to multiple access networks (e.g., Wi-Fi, LTE, and 5G) simultaneously. Recent advancements facilitate seamless integration of these connections below the transport layer, enhancing the experience for apps that lack inherent multi-path support. This optimization hinges on dynamically determining the traffic distribution across networks for each device, a process referred to as \textit{multi-access traffic splitting}. This paper introduces \textit{NetworkGym}, a high-fidelity network environment simulator that facilitates generating multiple network traffic flows and multi-access traffic splitting. This simulator facilitates training and evaluating different RL-based solutions for the multi-access traffic splitting problem. Our initial explorations demonstrate that the majority of existing state-of-the-art offline RL algorithms (e.g. CQL) fail to outperform certain hand-crafted heuristic policies on average. This illustrates the urgent need to evaluate offline RL algorithms against a broader range of benchmarks, rather than relying solely on popular ones such as D4RL. We also propose an extension to the TD3+BC algorithm, named Pessimistic TD3 (PTD3), and demonstrate that it outperforms many state-of-the-art offline RL algorithms. PTD3's behavioral constraint mechanism, which relies on value-function pessimism, is theoretically motivated and relatively simple to implement. 

We study offline multitask representation learning in reinforcement learning (RL), where a learner is provided with an offline dataset from different tasks that share a common representation and is tasked to learn the shared representation. We theoretically investigate offline multitask low-rank RL, and propose a new algorithm called MORL for offline multitask representation learning. Furthermore, we examine downstream RL in reward-free, offline and online scenarios, where a new task is introduced to the agent that shares the same representation as the upstream offline tasks. Our theoretical results demonstrate the benefits of using the learned representation from the upstream offline task instead of directly learning the representation of the low-rank model.