Recent XAI studies have investigated what constitutes a \textit{good} explanation in AI-assisted decision-making. Despite the widely accepted human-friendly properties of explanations, such as contrastive and selective, existing studies have yielded inconsistent findings. To address these gaps, our study focuses on the cognitive dimensions of explanation evaluation, by evaluating six explanations with different contrastive strategies and information selectivity and scrutinizing factors behind their valuation process. Our analysis results find that contrastive explanations are not the most preferable or understandable in general; Rather, different contrastive and selective explanations were appreciated to a different extent based on who they are, when, how, and what to explain -- with different level of cognitive load and engagement and sociotechnical contexts. Given these findings, we call for a nuanced view of explanation strategies, with implications for designing AI interfaces to accommodate individual and contextual differences in AI-assisted decision-making.

LLMs have emerged as a promising tool for assisting individuals in diverse text-generation tasks, including job-related texts. However, LLM-generated answers have been increasingly found to exhibit gender bias. This study evaluates three LLMs (GPT-3.5, GPT-4, Claude) to conduct a multifaceted audit of LLM-generated interview responses across models, question types, and jobs, and their alignment with two gender stereotypes. Our findings reveal that gender bias is consistent, and closely aligned with gender stereotypes and the dominance of jobs. Overall, this study contributes to the systematic examination of gender bias in LLM-generated interview responses, highlighting the need for a mindful approach to mitigate such biases in related applications.

In schools, teachers play a multitude of roles, serving as educators, counselors, decision-makers, and members of the school community. With recent advances in artificial intelligence (AI), there is increasing discussion about how AI can assist, complement, and collaborate with teachers. To pave the way for better teacher-AI complementary relationships in schools, our study aims to expand the discourse on teacher-AI complementarity by seeking educators' perspectives on the potential strengths and limitations of AI across a spectrum of responsibilities. Through a mixed method using a survey with 100 elementary school teachers in South Korea and in-depth interviews with 12 teachers, our findings indicate that teachers anticipate AI's potential to complement human teachers by automating administrative tasks and enhancing personalized learning through advanced intelligence. Interestingly, the deficit of AI's socio-emotional capabilities has been perceived as both challenges and opportunities. Overall, our study demonstrates the nuanced perception of teachers and different levels of expectations over their roles, challenging the need for decisions about AI adoption tailored to educators' preferences and concerns.

Despite the benefits of personalizing items and information tailored to users' needs, it has been found that recommender systems tend to introduce biases that favor popular items or certain categories of items, and dominant user groups. In this study, we aim to characterize the systematic errors of a recommendation system and how they manifest in various accountability issues, such as stereotypes, biases, and miscalibration. We propose a unified framework that distinguishes the sources of prediction errors into a set of key measures that quantify the various types of system-induced effects, both at the individual and collective levels. Based on our measuring framework, we examine the most widely adopted algorithms in the context of movie recommendation. Our research reveals three important findings: (1) Differences between algorithms: recommendations generated by simpler algorithms tend to be more stereotypical but less biased than those generated by more complex algorithms. (2) Disparate impact on groups and individuals: system-induced biases and stereotypes have a disproportionate effect on atypical users and minority groups (e.g., women and older users). (3) Mitigation opportunity: using structural equation modeling, we identify the interactions between user characteristics (typicality and diversity), system-induced effects, and miscalibration. We further investigate the possibility of mitigating system-induced effects by oversampling underrepresented groups and individuals, which was found to be effective in reducing stereotypes and improving recommendation quality. Our research is the first systematic examination of not only system-induced effects and miscalibration but also the stereotyping issue in recommender systems.

The escalating food insecurity in Africa, caused by factors such as war, climate change, and poverty, demonstrates the critical need for advanced early warning systems. Traditional methodologies, relying on expert-curated data encompassing climate, geography, and social disturbances, often fall short due to data limitations, hindering comprehensive analysis and potential discovery of new predictive factors. To address this, this paper introduces "HungerGist", a multi-task deep learning model utilizing news texts and NLP techniques. Using a corpus of over 53,000 news articles from nine African countries over four years, we demonstrate that our model, trained solely on news data, outperforms the baseline method trained on both traditional risk factors and human-curated keywords. In addition, our method has the ability to detect critical texts that contain interpretable signals known as "gists." Moreover, our examination of these gists indicates that this approach has the potential to reveal latent factors that would otherwise remain concealed in unstructured texts.

Big data and machine learning tools have jointly empowered humans in making data-driven decisions. However, many of them capture empirical associations that might be spurious due to confounding factors and subgroup heterogeneity. The famous Simpson's paradox is such a phenomenon where aggregated and subgroup-level associations contradict with each other, causing cognitive confusions and difficulty in making adequate interpretations and decisions. Existing tools provide little insights for humans to locate, reason about, and prevent pitfalls of spurious association in practice. We propose VISPUR, a visual analytic system that provides a causal analysis framework and a human-centric workflow for tackling spurious associations. These include a CONFOUNDER DASHBOARD, which can automatically identify possible confounding factors, and a SUBGROUP VIEWER, which allows for the visualization and comparison of diverse subgroup patterns that likely or potentially result in a misinterpretation of causality. Additionally, we propose a REASONING STORYBOARD, which uses a flow-based approach to illustrate paradoxical phenomena, as well as an interactive DECISION DIAGNOSIS panel that helps ensure accountable decision-making. Through an expert interview and a controlled user experiment, our qualitative and quantitative results demonstrate that the proposed "de-paradox" workflow and the designed visual analytic system are effective in helping human users to identify and understand spurious associations, as well as to make accountable causal decisions.

With the rise of big data, artificial intelligence (AI), and data mining techniques, group analysis has increasingly become a powerful tool in many applications, ranging from policy-making, direct marketing, education, to healthcare. For example, an important analysis strategy is group profiling, which extracts and describes the characteristics of groups of people [40]; it has been commonly used for customized recommendations to overcome sparse and missing personal data [25]. The same strategy is also used for mining social media, educational, and healthcare data to understand the shared characteristics of online communities or student/patient cohorts [15, 51, 100]. While it may help to support public and private services or product creations that are better tailored to different communities, group profiles resulted from mathematical inference are typically not valid for every individual regarded as a member in the group (this is known as non-distributive group profiles) [40]. The shared group characteristics extracted from data can have social ramifications such as stereotyping, stigmatization, or lead to pernicious consequences in decision making because individuals might be judged by group characteristics they do not posses [24, 56, 58].

Classification models learn to generalize the associations between data samples and their target classes. However, researchers have increasingly observed that machine learning practice easily leads to systematic errors in AI applications, a phenomenon referred to as AI blindspots. Such blindspots arise when a model is trained with training samples (e.g., cat/dog classification) where important patterns (e.g., black cats) are missing or periphery/undesirable patterns (e.g., dogs with grass background) are misleading towards a certain class. Even more sophisticated techniques cannot guarantee to capture, reason about, and prevent the spurious associations. In this work, we propose ESCAPE, a visual analytic system that promotes a human-in-the-loop workflow for countering systematic errors. By allowing human users to easily inspect spurious associations, the system facilitates users to spontaneously recognize concepts associated misclassifications and evaluate mitigation strategies that can reduce biased associations. We also propose two statistical approaches, relative concept association to better quantify the associations between a concept and instances, and debias method to mitigate spurious associations. We demonstrate the utility of our proposed ESCAPE system and statistical measures through extensive evaluation including quantitative experiments, usage scenarios, expert interviews, and controlled user experiments.