Hua Shen

Tongshuang Wu, Hua Shen, Daniel S. Weld et al. · cmu, microsoft-research

The in-context learning capabilities of LLMs like GPT-3 allow annotators to customize an LLM to their specific tasks with a small number of examples. However, users tend to include only the most obvious patterns when crafting examples, resulting in underspecified in-context functions that fall short on unseen cases. Further, it is hard to know when "enough" examples have been included even for known patterns. In this work, we present ScatterShot, an interactive system for building high-quality demonstration sets for in-context learning. ScatterShot iteratively slices unlabeled data into task-specific patterns, samples informative inputs from underexplored or not-yet-saturated slices in an active learning manner, and helps users label more efficiently with the help of an LLM and the current example set. In simulation studies on two text perturbation scenarios, ScatterShot sampling improves the resulting few-shot functions by 4-5 percentage points over random sampling, with less variance as more examples are added. In a user study, ScatterShot greatly helps users in covering different patterns in the input space and labeling in-context examples more efficiently, resulting in better in-context learning and less user effort.

Kai-Wei Chang, Yu-Kai Wang, Hua Shen et al. · meta-ai, mit

Prompt tuning is a technology that tunes a small set of parameters to steer a pre-trained language model (LM) to directly generate the output for downstream tasks. Recently, prompt tuning has demonstrated its storage and computation efficiency in both natural language processing (NLP) and speech processing fields. These advantages have also revealed prompt tuning as a candidate approach to serving pre-trained LM for multiple tasks in a unified manner. For speech processing, SpeechPrompt shows its high parameter efficiency and competitive performance on a few speech classification tasks. However, whether SpeechPrompt is capable of serving a large number of tasks is unanswered. In this work, we propose SpeechPrompt v2, a prompt tuning framework capable of performing a wide variety of speech classification tasks, covering multiple languages and prosody-related tasks. The experiment result shows that SpeechPrompt v2 achieves performance on par with prior works with less than 0.15M trainable parameters in a unified framework.

Adaku Uchendu, Jooyoung Lee, Hua Shen et al.

Advances in Large Language Models (e.g., GPT-4, LLaMA) have improved the generation of coherent sentences resembling human writing on a large scale, resulting in the creation of so-called deepfake texts. However, this progress poses security and privacy concerns, necessitating effective solutions for distinguishing deepfake texts from human-written ones. Although prior works studied humans' ability to detect deepfake texts, none has examined whether "collaboration" among humans improves the detection of deepfake texts. In this study, to address this gap of understanding on deepfake texts, we conducted experiments with two groups: (1) nonexpert individuals from the AMT platform and (2) writing experts from the Upwork platform. The results demonstrate that collaboration among humans can potentially improve the detection of deepfake texts for both groups, increasing detection accuracies by 6.36% for non-experts and 12.76% for experts, respectively, compared to individuals' detection accuracies. We further analyze the explanations that humans used for detecting a piece of text as deepfake text, and find that the strongest indicator of deepfake texts is their lack of coherence and consistency. Our study provides useful insights for future tools and framework designs to facilitate the collaborative human detection of deepfake texts. The experiment datasets and AMT implementations are available at: https://github.com/huashen218/llm-deepfake-human-study.git

Kai-Wei Chang, Haibin Wu, Yu-Kai Wang et al. · meta-ai, mit

Prompting has become a practical method for utilizing pre-trained language models (LMs). This approach offers several advantages. It allows an LM to adapt to new tasks with minimal training and parameter updates, thus achieving efficiency in both storage and computation. Additionally, prompting modifies only the LM's inputs and harnesses the generative capabilities of language models to address various downstream tasks in a unified manner. This significantly reduces the need for human labor in designing task-specific models. These advantages become even more evident as the number of tasks served by the LM scales up. Motivated by the strengths of prompting, we are the first to explore the potential of prompting speech LMs in the domain of speech processing. Recently, there has been a growing interest in converting speech into discrete units for language modeling. Our pioneer research demonstrates that these quantized speech units are highly versatile within our unified prompting framework. Not only can they serve as class labels, but they also contain rich phonetic information that can be re-synthesized back into speech signals for speech generation tasks. Specifically, we reformulate speech processing tasks into speech-to-unit generation tasks. As a result, we can seamlessly integrate tasks such as speech classification, sequence generation, and speech generation within a single, unified prompting framework. The experiment results show that the prompting method can achieve competitive performance compared to the strong fine-tuning method based on self-supervised learning models with a similar number of trainable parameters. The prompting method also shows promising results in the few-shot setting. Moreover, with the advanced speech LMs coming into the stage, the proposed prompting framework attains great potential.

Hua Shen, Tongshuang Wu, Wenbo Guo et al. · cmu

Existing self-explaining models typically favor extracting the shortest possible rationales - snippets of an input text "responsible for" corresponding output - to explain the model prediction, with the assumption that shorter rationales are more intuitive to humans. However, this assumption has yet to be validated. Is the shortest rationale indeed the most human-understandable? To answer this question, we design a self-explaining model, LimitedInk, which allows users to extract rationales at any target length. Compared to existing baselines, LimitedInk achieves compatible end-task performance and human-annotated rationale agreement, making it a suitable representation of the recent class of self-explaining models. We use LimitedInk to conduct a user study on the impact of rationale length, where we ask human judges to predict the sentiment label of documents based only on LimitedInk-generated rationales with different lengths. We show rationales that are too short do not help humans predict labels better than randomly masked text, suggesting the need for more careful design of the best human rationales.

Hua Shen, Tongshuang Wu · cmu

A surge of advances in language models (LMs) has led to significant interest in using LMs to build co-writing systems, in which humans and LMs interactively contribute to a shared writing artifact. However, there is a lack of studies assessing co-writing systems in interactive settings. We propose a human-centered evaluation framework, Parachute, for interactive co-writing systems. Parachute showcases an integrative view of interaction evaluation, where each evaluation aspect consists of categorized practical metrics. Furthermore, we present Parachute with a use case to demonstrate how to evaluate and compare co-writing systems using Parachute.

Jonathan Ivey, Shivani Kumar, Jiayu Liu et al.

Studying and building datasets for dialogue tasks is both expensive and time-consuming due to the need to recruit, train, and collect data from study participants. In response, much recent work has sought to use large language models (LLMs) to simulate both human-human and human-LLM interactions, as they have been shown to generate convincingly human-like text in many settings. However, to what extent do LLM-based simulations \textit{actually} reflect human dialogues? In this work, we answer this question by generating a large-scale dataset of 100,000 paired LLM-LLM and human-LLM dialogues from the WildChat dataset and quantifying how well the LLM simulations align with their human counterparts. Overall, we find relatively low alignment between simulations and human interactions, demonstrating a systematic divergence along the multiple textual properties, including style and content. Further, in comparisons of English, Chinese, and Russian dialogues, we find that models perform similarly. Our results suggest that LLMs generally perform better when the human themself writes in a way that is more similar to the LLM's own style.

Binfeng Xu, Xukun Liu, Hua Shen et al.

Augmented Language Models (ALMs) empower large language models with the ability to use tools, transforming them into intelligent agents for real-world interactions. However, most existing frameworks for ALMs, to varying degrees, are deficient in the following critical features: flexible customization, collaborative democratization, and holistic evaluation. We present gentopia, an ALM framework enabling flexible customization of agents through simple configurations, seamlessly integrating various language models, task formats, prompting modules, and plugins into a unified paradigm. Furthermore, we establish gentpool, a public platform enabling the registration and sharing of user-customized agents. Agents registered in gentpool are composable such that they can be assembled together for agent collaboration, advancing the democratization of artificial intelligence. To ensure high-quality agents, gentbench, an integral component of gentpool, is designed to thoroughly evaluate user-customized agents across diverse aspects such as safety, robustness, efficiency, etc. We release gentopia on Github and will continuously move forward.

Sushrita Rakshit, Hanwen Zhang, Hua Shen

Large language models (LLMs) are often evaluated based on their stated values, yet these do not reliably translate into their actions, a discrepancy termed "value-action gap." In this work, we argue that this gap persists even under explicit reasoning, revealing a deeper failure mode we call "Pseudo-Deliberation": the appearance of principled reasoning without corresponding behavioral alignment. To study this systematically, we introduce VALDI, a framework for measuring alignment between stated values and generated dialogue. VALDI includes 4,941 human-centered scenarios across five domains, three tasks that elicit value articulation, reasoning, and action, and five metrics for quantifying value adherence. Across both proprietary and open-source LLMs, we observe consistent misalignment between expressed values and downstream dialogues. To investigate intervention strategies, we propose VIVALDI, a multi-agent value auditor that intervenes at different stages of generation.

Qianou Ma, Weirui Peng, Chenyang Yang et al.

Prompting LLMs for complex tasks (e.g., building a trip advisor chatbot) needs humans to clearly articulate customized requirements (e.g., "start the response with a tl;dr"). However, existing prompt engineering instructions often lack focused training on requirement articulation and instead tend to emphasize increasingly automatable strategies (e.g., tricks like adding role-plays and "think step-by-step"). To address the gap, we introduce Requirement-Oriented Prompt Engineering (ROPE), a paradigm that focuses human attention on generating clear, complete requirements during prompting. We implement ROPE through an assessment and training suite that provides deliberate practice with LLM-generated feedback. In a randomized controlled experiment with 30 novices, ROPE significantly outperforms conventional prompt engineering training (20% vs. 1% gains), a gap that automatic prompt optimization cannot close. Furthermore, we demonstrate a direct correlation between the quality of input requirements and LLM outputs. Our work paves the way to empower more end-users to build complex LLM applications.

Hua Shen, Tiffany Knearem, Divy Thakkar et al.

The rapid integration of generative AI into everyday life underscores the need to move beyond unidirectional alignment models that only adapt AI to human values. This workshop focuses on bidirectional human-AI alignment, a dynamic, reciprocal process where humans and AI co-adapt through interaction, evaluation, and value-centered design. Building on our past CHI 2025 BiAlign SIG and ICLR 2025 Workshop, this workshop will bring together interdisciplinary researchers from HCI, AI, social sciences and more domains to advance value-centered AI and reciprocal human-AI collaboration. We focus on embedding human and societal values into alignment research, emphasizing not only steering AI toward human values but also enabling humans to critically engage with and evolve alongside AI systems. Through talks, interdisciplinary discussions, and collaborative activities, participants will explore methods for interactive alignment, frameworks for societal impact evaluation, and strategies for alignment in dynamic contexts. This workshop aims to bridge the disciplines' gaps and establish a shared agenda for responsible, reciprocal human-AI futures.

Hua Shen, Tiffany Knearem, Reshmi Ghosh et al.

As AI systems become more advanced, ensuring their alignment with a diverse range of individuals and societal values becomes increasingly critical. But how can we capture fundamental human values and assess the degree to which AI systems align with them? We introduce ValueCompass, a framework of fundamental values, grounded in psychological theory and a systematic review, to identify and evaluate human-AI alignment. We apply ValueCompass to measure the value alignment of humans and large language models (LLMs) across four real-world scenarios: collaborative writing, education, public sectors, and healthcare. Our findings reveal concerning misalignments between humans and LLMs, such as humans frequently endorse values like "National Security" which were largely rejected by LLMs. We also observe that values differ across scenarios, highlighting the need for context-aware AI alignment strategies. This work provides valuable insights into the design space of human-AI alignment, laying the foundations for developing AI systems that responsibly reflect societal values and ethics.

Xinyu Pi, Qisen Yang, Chuong Nguyen et al.

LLMs are increasingly used to support qualitative research, yet existing systems produce outputs that vary widely--from trace-faithful summaries to theory-mediated explanations and system models. To make these differences explicit, we introduce a 4$\times$4 landscape crossing four levels of meaning-making (descriptive, categorical, interpretive, theoretical) with four levels of modeling (static structure, stages/timelines, causal pathways, feedback dynamics). Applying the landscape to prior LLM-based automation highlights a strong skew toward low-level meaning and low-commitment representations, with few reliable attempts at interpretive/theoretical inference or dynamical modeling. Based on the revealed gap, we outline an agenda for applying and building LLM-systems that make their interpretive and modeling commitments explicit, selectable, and governable.

Mina Lee, Katy Ilonka Gero, John Joon Young Chung et al. · allen-ai, deepmind

In our era of rapid technological advancement, the research landscape for writing assistants has become increasingly fragmented across various research communities. We seek to address this challenge by proposing a design space as a structured way to examine and explore the multidimensional space of intelligent and interactive writing assistants. Through a large community collaboration, we explore five aspects of writing assistants: task, user, technology, interaction, and ecosystem. Within each aspect, we define dimensions (i.e., fundamental components of an aspect) and codes (i.e., potential options for each dimension) by systematically reviewing 115 papers. Our design space aims to offer researchers and designers a practical tool to navigate, comprehend, and compare the various possibilities of writing assistants, and aid in the envisioning and design of new writing assistants.

Joshua Ashkinaze, Hua Shen, Sai Avula et al.

We introduce the Deep Value Benchmark (DVB), an evaluation framework that directly tests whether large language models (LLMs) learn fundamental human values or merely surface-level preferences. This distinction is critical for AI alignment: Systems that capture deeper values are likely to generalize human intentions robustly, while those that capture only superficial patterns in preference data risk producing misaligned behavior. The DVB uses a novel experimental design with controlled confounding between deep values (e.g., moral principles) and shallow features (e.g., superficial attributes). In the training phase, we expose LLMs to human preference data with deliberately correlated deep and shallow features -- for instance, where a user consistently prefers (non-maleficence, formal language) options over (justice, informal language) alternatives. The testing phase then breaks these correlations, presenting choices between (justice, formal language) and (non-maleficence, informal language) options. This design allows us to precisely measure a model's Deep Value Generalization Rate (DVGR) -- the probability of generalizing based on the underlying value rather than the shallow feature. Across 9 different models, the average DVGR is just 0.30. All models generalize deep values less than chance. Larger models have a (slightly) lower DVGR than smaller models. We are releasing our dataset, which was subject to three separate human validation experiments. DVB provides an interpretable measure of a core feature of alignment.

Yike Shi, Qing Xiao, Qing Hu et al.

Large language models can influence users through conversation, creating new forms of dark patterns that differ from traditional UX dark patterns. We define LLM dark patterns as manipulative or deceptive behaviors enacted in dialogue. Drawing on prior work and AI incident reports, we outline a diverse set of categories with real-world examples. Using them, we conducted a scenario-based study where participants (N=34) compared manipulative and neutral LLM responses. Our results reveal that recognition of LLM dark patterns often hinged on conversational cues such as exaggerated agreement, biased framing, or privacy intrusions, but these behaviors were also sometimes normalized as ordinary assistance. Users' perceptions of these dark patterns shaped how they respond to them. Responsibilities for these behaviors were also attributed in different ways, with participants assigning it to companies and developers, the model itself, or to users. We conclude with implications for design, advocacy, and governance to safeguard user autonomy.

Tuc Nguyen, James Michels, Hua Shen et al.

In Explainable AI (XAI), counterfactual explanations (CEs) are a well-studied method to communicate feature relevance through contrastive reasoning of "what if" to explain AI models' predictions. However, they only focus on important (i.e., relevant) features and largely disregard less important (i.e., irrelevant) ones. Such irrelevant features can be crucial in many applications, especially when users need to ensure that an AI model's decisions are not affected or biased against specific attributes such as gender, race, religion, or political affiliation. To address this gap, the concept of alterfactual explanations (AEs) has been proposed. AEs explore an alternative reality of "no matter what", where irrelevant features are substituted with alternative features (e.g., "republicans" -> "democrats") within the same attribute (e.g., "politics") while maintaining a similar prediction output. This serves to validate whether AI model predictions are influenced by the specified attributes. Despite the promise of AEs, there is a lack of computational approaches to systematically generate them, particularly in the text domain, where creating AEs for AI text classifiers presents unique challenges. This paper addresses this challenge by formulating AE generation as an optimization problem and introducing MoMatterXAI, a novel algorithm that generates AEs for text classification tasks. Our approach achieves high fidelity of up to 95% while preserving context similarity of over 90% across multiple models and datasets. A human study further validates the effectiveness of AEs in explaining AI text classifiers to end users. All codes will be publicly available.

Hua Shen, Nicholas Clark, Tanushree Mitra

Existing research primarily evaluates the values of LLMs by examining their stated inclinations towards specific values. However, the "Value-Action Gap," a phenomenon rooted in environmental and social psychology, reveals discrepancies between individuals' stated values and their actions in real-world contexts. To what extent do LLMs exhibit a similar gap between their stated values and their actions informed by those values? This study introduces ValueActionLens, an evaluation framework to assess the alignment between LLMs' stated values and their value-informed actions. The framework encompasses the generation of a dataset comprising 14.8k value-informed actions across twelve cultures and eleven social topics, and two tasks to evaluate how well LLMs' stated value inclinations and value-informed actions align across three different alignment measures. Extensive experiments reveal that the alignment between LLMs' stated values and actions is sub-optimal, varying significantly across scenarios and models. Analysis of misaligned results identifies potential harms from certain value-action gaps. To predict the value-action gaps, we also uncover that leveraging reasoned explanations improves performance. These findings underscore the risks of relying solely on the LLMs' stated values to predict their behaviors and emphasize the importance of context-aware evaluations of LLM values and value-action gaps.

Hua Shen

Artificial intelligence (AI) is transforming education, offering unprecedented opportunities to personalize learning, enhance assessment, and support educators. Yet these opportunities also introduce risks related to equity, privacy, and student autonomy. This chapter develops the concept of bidirectional human-AI alignment in education, emphasizing that trustworthy learning environments arise not only from embedding human values into AI systems but also from equipping teachers, students, and institutions with the skills to interpret, critique, and guide these technologies. Drawing on emerging research and practical case examples, we explore AI's evolution from support tool to collaborative partner, highlighting its impacts on teacher roles, student agency, and institutional governance. We propose actionable strategies for policymakers, developers, and educators to ensure that AI advances equity, transparency, and human flourishing rather than eroding them. By reframing AI adoption as an ongoing process of mutual adaptation, the chapter envisions a future in which humans and intelligent systems learn, innovate, and grow together.

Nicholas Clark, Hua Shen, Bill Howe et al.

LLMs increasingly serve as tools for knowledge acquisition, yet users cannot effectively specify how they want information presented. When users request that LLMs "cite reputable sources," "express appropriate uncertainty," or "include multiple perspectives," they discover that current interfaces provide no structured way to articulate these preferences. The result is prompt sharing folklore: community-specific copied prompts passed through trust relationships rather than based on measured efficacy. We propose the Epistemic Alignment Framework, a set of ten challenges in knowledge transmission derived from the philosophical literature of epistemology, concerning issues such as evidence quality assessment and calibration of testimonial reliance. The framework serves as a structured intermediary between user needs and system capabilities, creating a common vocabulary to bridge the gap between what users want and what systems deliver. Through a thematic analysis of custom prompts and personalization strategies shared on online communities where these issues are actively discussed, we find users develop elaborate workarounds to address each of the challenges. We then apply our framework to two prominent model providers, OpenAI and Anthropic, through content analysis of their documented policies and product features. Our analysis shows that while these providers have partially addressed the challenges we identified, they fail to establish adequate mechanisms for specifying epistemic preferences, lack transparency about how preferences are implemented, and offer no verification tools to confirm whether preferences were followed. For AI developers, the Epistemic Alignment Framework offers concrete guidance for supporting diverse approaches to knowledge; for users, it works toward information delivery that aligns with their specific needs rather than defaulting to one-size-fits-all approaches.

Jiajun Chen, Hua Shen

Existing work on value alignment typically characterizes value relations statically, ignoring how interventions - such as prompting, fine-tuning, or preference optimization - reshape the broader value system. We introduce the Value Alignment Tax (VAT), a framework that measures how alignment-induced changes propagate across interconnected values relative to achieved on-target gain. VAT captures the dynamics of value expression under alignment pressure. Using a controlled scenario-action dataset grounded in Schwartz value theory, we collect paired pre-post normative judgments and analyze alignment effects across models, values, and alignment strategies. Our results show that alignment often produces uneven, structured co-movement among values. These effects are invisible under conventional target-only evaluation, revealing systemic, process-level alignment risks and offering new insights into the dynamics of value alignment in LLMs.

Jinnuo Liu, Chuke Liu, Hua Shen

Multi-agent large language model (LLM) systems increasingly consist of agents that observe and respond to one another's outputs. While value alignment is typically evaluated for isolated models, how value perturbations propagate through agent interactions remains poorly understood. We present ValueFlow, a perturbation-based evaluation framework for measuring and analyzing value drift in multi-agent systems. ValueFlow introduces a 56-value evaluation dataset derived from the Schwartz Value Survey and quantifies agents' value orientations during interaction using an LLM-as-a-judge protocol. Building on this measurement layer, ValueFlow decomposes value drift into agent-level response behavior and system-level structural effects, operationalized by two metrics: beta-susceptibility, which measures an agent's sensitivity to perturbed peer signals, and system susceptibility (SS), which captures how node-level perturbations affect final system outputs. Experiments across multiple model backbones, prompt personas, value dimensions, and network structures show that susceptibility varies widely across values and is strongly shaped by structural topology.

Yuyang Jiang, Longjie Guo, Yuchen Wu et al.

Large language model (LLM)-powered chatbots are increasingly used for opinion exploration. Prior research examined how LLMs alter user views, yet little work extended beyond one-way influence to address how user input can affect LLM responses and how such bi-directional influence manifests throughout the multi-turn conversations. This study investigates this dynamic through 50 controversial-topic discussions with participants (N=266) across three conditions: static statements, standard chatbot, and personalized chatbot. Results show that human opinions barely shifted, while LLM outputs changed more substantially, narrowing the gap between human and LLM stance. Personalization amplified these shifts in both directions compared to the standard setting. Analysis of multi-turn conversations further revealed that exchanges involving participants' personal stories were most likely to trigger stance changes for both humans and LLMs. Our work highlights the risk of over-alignment in human-LLM interaction and the need for careful design of personalized chatbots to more thoughtfully and stably align with users.

Hua Shen, Tiffany Knearem, Reshmi Ghosh et al.

Recent advances in general-purpose AI underscore the urgent need to align AI systems with human goals and values. Yet, the lack of a clear, shared understanding of what constitutes "alignment" limits meaningful progress and cross-disciplinary collaboration. In this position paper, we argue that the research community should explicitly define and critically reflect on "alignment" to account for the bidirectional and dynamic relationship between humans and AI. Through a systematic review of over 400 papers spanning HCI, NLP, ML, and more, we examine how alignment is currently defined and operationalized. Building on this analysis, we introduce the Bidirectional Human-AI Alignment framework, which not only incorporates traditional efforts to align AI with human values but also introduces the critical, underexplored dimension of aligning humans with AI -- supporting cognitive, behavioral, and societal adaptation to rapidly advancing AI technologies. Our findings reveal significant gaps in current literature, especially in long-term interaction design, human value modeling, and mutual understanding. We conclude with three central challenges and actionable recommendations to guide future research toward more nuanced, reciprocal, and human-AI alignment approaches.

Hua Shen, Vicky Zayats, Johann C. Rocholl et al.

Current disfluency detection models focus on individual utterances each from a single speaker. However, numerous discontinuity phenomena in spoken conversational transcripts occur across multiple turns, hampering human readability and the performance of downstream NLP tasks. This study addresses these phenomena by proposing an innovative Multi-Turn Cleanup task for spoken conversational transcripts and collecting a new dataset, MultiTurnCleanup1. We design a data labeling schema to collect the high-quality dataset and provide extensive data analysis. Furthermore, we leverage two modeling approaches for experimental evaluation as benchmarks for future research.

Hua Shen, Chieh-Yang Huang, Tongshuang Wu et al.

Despite a surge collection of XAI methods, users still struggle to obtain required AI explanations. Previous research suggests chatbots as dynamic solutions, but the effective design of conversational XAI agents for practical human needs remains under-explored. This paper focuses on Conversational XAI for AI-assisted scientific writing tasks. Drawing from human linguistic theories and formative studies, we identify four design rationales: "multifaceted", "controllability", "mix-initiative", "context-aware drill-down". We incorporate them into an interactive prototype, ConvXAI, which facilitates heterogeneous AI explanations for scientific writing through dialogue. In two studies with 21 users, ConvXAI outperforms a GUI-based baseline on improving human-perceived understanding and writing improvement. The paper further discusses the practical human usage patterns in interacting with ConvXAI for scientific co-writing.

Hua Shen, Yuguang Yang, Guoli Sun et al.

Modern speaker verification models use deep neural networks to encode utterance audio into discriminative embedding vectors. During the training process, these networks are typically optimized to differentiate arbitrary speakers. This learning process biases the learning of fine voice characteristics towards dominant demographic groups, which can lead to an unfair performance disparity across different groups. This is observed especially with underrepresented demographic groups sharing similar voice characteristics. In this work, we investigate the fairness of speaker verification models on controlled datasets with imbalanced gender distributions, providing direct evidence that model performance suffers for underrepresented groups. To mitigate this disparity we propose the group-adapted fusion network (GFN) architecture, a modular architecture based on group embedding adaptation and score fusion. We show that our method alleviates model unfairness by improving speaker verification both overall and for individual groups. Given imbalanced group representation in training, our proposed method achieves overall equal error rate (EER) reduction of 9.6% to 29.0% relative, reduces minority group EER by 13.7% to 18.6%, and results in 20.0% to 25.4% less EER disparity, compared to baselines. The approach is applicable to other types of training data skew in speaker recognition systems.

Hua Shen, Ting-Hao 'Kenneth' Huang

It is unclear if existing interpretations of deep neural network models respond effectively to the needs of users. This paper summarizes the common forms of explanations (such as feature attribution, decision rules, or probes) used in over 200 recent papers about natural language processing (NLP), and compares them against user questions collected in the XAI Question Bank. We found that although users are interested in explanations for the road not taken -- namely, why the model chose one result and not a well-defined, seemly similar legitimate counterpart -- most model interpretations cannot answer these questions.

Hua Shen, Ting-Hao Kenneth Huang

Explaining to users why automated systems make certain mistakes is important and challenging. Researchers have proposed ways to automatically produce interpretations for deep neural network models. However, it is unclear how useful these interpretations are in helping users figure out why they are getting an error. If an interpretation effectively explains to users how the underlying deep neural network model works, people who were presented with the interpretation should be better at predicting the model's outputs than those who were not. This paper presents an investigation on whether or not showing machine-generated visual interpretations helps users understand the incorrectly predicted labels produced by image classifiers. We showed the images and the correct labels to 150 online crowd workers and asked them to select the incorrectly predicted labels with or without showing them the machine-generated visual interpretations. The results demonstrated that displaying the visual interpretations did not increase, but rather decreased, the average guessing accuracy by roughly 10%.

Ren Pang, Hua Shen, Xinyang Zhang et al.

Despite their tremendous success in a range of domains, deep learning systems are inherently susceptible to two types of manipulations: adversarial inputs -- maliciously crafted samples that deceive target deep neural network (DNN) models, and poisoned models -- adversely forged DNNs that misbehave on pre-defined inputs. While prior work has intensively studied the two attack vectors in parallel, there is still a lack of understanding about their fundamental connections: what are the dynamic interactions between the two attack vectors? what are the implications of such interactions for optimizing existing attacks? what are the potential countermeasures against the enhanced attacks? Answering these key questions is crucial for assessing and mitigating the holistic vulnerabilities of DNNs deployed in realistic settings. Here we take a solid step towards this goal by conducting the first systematic study of the two attack vectors within a unified framework. Specifically, (i) we develop a new attack model that jointly optimizes adversarial inputs and poisoned models; (ii) with both analytical and empirical evidence, we reveal that there exist intriguing "mutual reinforcement" effects between the two attack vectors -- leveraging one vector significantly amplifies the effectiveness of the other; (iii) we demonstrate that such effects enable a large design spectrum for the adversary to enhance the existing attacks that exploit both vectors (e.g., backdoor attacks), such as maximizing the attack evasiveness with respect to various detection methods; (iv) finally, we discuss potential countermeasures against such optimized attacks and their technical challenges, pointing to several promising research directions.

Xinyang Zhang, Ningfei Wang, Hua Shen et al.

Providing explanations for deep neural network (DNN) models is crucial for their use in security-sensitive domains. A plethora of interpretation models have been proposed to help users understand the inner workings of DNNs: how does a DNN arrive at a specific decision for a given input? The improved interpretability is believed to offer a sense of security by involving human in the decision-making process. Yet, due to its data-driven nature, the interpretability itself is potentially susceptible to malicious manipulations, about which little is known thus far. Here we bridge this gap by conducting the first systematic study on the security of interpretable deep learning systems (IDLSes). We show that existing \imlses are highly vulnerable to adversarial manipulations. Specifically, we present ADV^2, a new class of attacks that generate adversarial inputs not only misleading target DNNs but also deceiving their coupled interpretation models. Through empirical evaluation against four major types of IDLSes on benchmark datasets and in security-critical applications (e.g., skin cancer diagnosis), we demonstrate that with ADV^2 the adversary is able to arbitrarily designate an input's prediction and interpretation. Further, with both analytical and empirical evidence, we identify the prediction-interpretation gap as one root cause of this vulnerability -- a DNN and its interpretation model are often misaligned, resulting in the possibility of exploiting both models simultaneously. Finally, we explore potential countermeasures against ADV^2, including leveraging its low transferability and incorporating it in an adversarial training framework. Our findings shed light on designing and operating IDLSes in a more secure and informative fashion, leading to several promising research directions.