Lydia B. Chilton

Sharon Levy, Emily Allaway, Melanie Subbiah et al.

Understanding what constitutes safe text is an important issue in natural language processing and can often prevent the deployment of models deemed harmful and unsafe. One such type of safety that has been scarcely studied is commonsense physical safety, i.e. text that is not explicitly violent and requires additional commonsense knowledge to comprehend that it leads to physical harm. We create the first benchmark dataset, SafeText, comprising real-life scenarios with paired safe and physically unsafe pieces of advice. We utilize SafeText to empirically study commonsense physical safety across various models designed for text generation and commonsense reasoning tasks. We find that state-of-the-art large language models are susceptible to the generation of unsafe text and have difficulty rejecting unsafe advice. As a result, we argue for further studies of safety and the assessment of commonsense physical safety in models before release.

Sitong Wang, Samia Menon, Tao Long et al.

Short videos on social media are the dominant way young people consume content. News outlets aim to reach audiences through news reels -- short videos conveying news -- but struggle to translate traditional journalistic formats into short, entertaining videos. To translate news into social media reels, we support journalists in reframing the narrative. In literature, narrative framing is a high-level structure that shapes the overall presentation of a story. We identified three narrative framings for reels that adapt social media norms but preserve news value, each with a different balance of information and entertainment. We introduce ReelFramer, a human-AI co-creative system that helps journalists translate print articles into scripts and storyboards. ReelFramer supports exploring multiple narrative framings to find one appropriate to the story. AI suggests foundational narrative details, including characters, plot, setting, and key information. ReelFramer also supports visual framing; AI suggests character and visual detail designs before generating a full storyboard. Our studies show that narrative framing introduces the necessary diversity to translate various articles into reels, and establishing foundational details helps generate scripts that are more relevant and coherent. We also discuss the benefits of using narrative framing and foundational details in content retargeting.

Vivian Liu, Tao Long, Nathan Raw et al.

Visuals can enhance our experience of music, owing to the way they can amplify the emotions and messages conveyed within it. However, creating music visualization is a complex, time-consuming, and resource-intensive process. We introduce Generative Disco, a generative AI system that helps generate music visualizations with large language models and text-to-video generation. The system helps users visualize music in intervals by finding prompts to describe the images that intervals start and end on and interpolating between them to the beat of the music. We introduce design patterns for improving these generated videos: transitions, which express shifts in color, time, subject, or style, and holds, which help focus the video on subjects. A study with professionals showed that transitions and holds were a highly expressive framework that enabled them to build coherent visual narratives. We conclude on the generalizability of these patterns and the potential of generated video for creative professionals.

Sitong Wang, Samia Menon, Dingzeyu Li et al.

Creative and communicative work is often underpinned by implicit structures, such as the Hero's Journey in storytelling, design patterns in software, or chord progressions in music. People often learn these structures from examples - a process known as schema induction. However, because schemas are abstract and implicit, they are difficult to discover: shared structural patterns are obscured by surface-level variation, and balancing generality with specificity is challenging. We present Schemex, an interactive AI workflow that systematically supports schema induction by decomposing it into three tractable stages: clustering examples, abstracting candidate schemas, and contrastively refining them by generating new instances and comparing against originals. Studies show that Schemex produces more actionable schemas than a frontier baseline without sacrificing generalizability, with participants uncovering deep and nuanced structural patterns. We also discuss design implications for the cognitive role of interactive process in structure discovery.

Melanie Subbiah, Faisal Ladhak, Akankshya Mishra et al.

Human evaluation has been the gold standard for checking faithfulness in abstractive summarization. However, with a challenging source domain like narrative, multiple annotators can agree a summary is faithful, while missing details that are obvious errors only once pointed out. We therefore introduce a new dataset, STORYSUMM, comprising LLM summaries of short stories with localized faithfulness labels and error explanations. This benchmark is for evaluation methods, testing whether a given method can detect challenging inconsistencies. Using this dataset, we first show that any one human annotation protocol is likely to miss inconsistencies, and we advocate for pursuing a range of methods when establishing ground truth for a summarization dataset. We finally test recent automatic metrics and find that none of them achieve more than 70% balanced accuracy on this task, demonstrating that it is a challenging benchmark for future work in faithfulness evaluation.

Tao Long, Lydia B. Chilton

Advances in voice technology and voice user interfaces (VUIs) -- such as Alexa, Siri, and Google Home -- have opened up the potential for many new types of interaction. However, despite the potential of these devices reflected by the growing market and body of VUI research, there is a lingering sense that the technology is still underused. In this paper, we conducted a systematic literature review of 35 papers to identify and synthesize 127 VUI design guidelines into five themes. Additionally, we conducted semi-structured interviews with 15 smart speaker users to understand their use and non-use of the technology. From the interviews, we distill four design challenges that contribute the most to non-use. Based on their (non-)use, we identify four opportunity spaces for designers to explore such as focusing on information support while multitasking (cooking, driving, childcare, etc), incorporating users' mental models for smart speakers, and integrating calm design principles.

Karthik Sreedhar, Aryan Kaul, Lydia B. Chilton

Customization has long been a central goal in interactive systems, yet prior work shows that end-user tailoring occurs infrequently and is often confined to initial setup or moments of breakdown. Recent advances in generative AI suggest that highly malleable systems-where users can modify system behavior through natural language-are now technically feasible. However, it remains unclear how such malleability is used in practice: What kinds of customizations do users create, when do they choose to customize, and how do these modifications shape their experience of everyday tools? We present a design probe that uses a conversationally customizable email system as an instrument to study how users create and refine functionality within everyday tools. The system allows users to iteratively modify their inbox by restructuring categories, introducing interface elements, and authoring new workflow behaviors directly through natural language interaction. We study how participants create, refine, and use these features over several days within their own email workflows. We find that users' customizations are often grounded in existing patterns, which they adapt and specialize to fit their needs, rather than generating entirely novel functionality. Malleability changes how users engage with their inbox, shifting it from a fixed interface to a flexible data layer shaped through user-authored features. At the same time, customization introduces new forms of risk, including mis-specified behavior, unintended filtering, and uncertainty around outcomes, which users manage through ongoing oversight and refinement. These findings highlight how conversational customization becomes embedded within everyday interaction, and point toward the need for systems that support iterative refinement, visibility into behavior, and safe experimentation as users shape their own tools.

Jenny Ma, Sitong Wang, Joshua H. Kung et al.

Rules files (e.g., AGENTS.md, CLAUDE.md) are the primary mechanism for human-agent alignment when developers vibe code. However, they remain passive: it is not immediately apparent when rules are being used or followed, or how to improve them. To transform rules from passive text into active controls, we introduce ZORO, an interactive interface that integrates directly with a coding agent and anchors rules to every step of the coding process. After an agent generates an initial plan, ZORO enriches the plan with rules, enforces the rules during implementation by requiring the agent prove that each rule was followed, and allows users to provide in-situ feedback when they are unsatisfied with a rule application to evolve the ruleset. A technical evaluation shows that coding agents follow rules more with ZORO than without. A user study demonstrates a change in people's behavior and cognitive strategies when rules are at the forefront of vibe coding. We discuss how making rules active in agentic systems unlocks broader opportunities for human-agent alignment in coding settings and beyond.

Tao Long, Katy Ilonka Gero, Lydia B. Chilton

Generative AI brings novel and impressive abilities to help people in everyday tasks. There are many AI workflows that solve real and complex problems by chaining AI outputs together with human interaction. Although there is an undeniable lure of AI, it is uncertain how useful generative AI workflows are after the novelty wears off. Additionally, workflows built with generative AI have the potential to be easily customized to fit users' individual needs, but do users take advantage of this? We conducted a three-week longitudinal study with 12 users to understand the familiarization and customization of generative AI tools for science communication. Our study revealed that there exists a familiarization phase, during which users were exploring the novel capabilities of the workflow and discovering which aspects they found useful. After this phase, users understood the workflow and were able to anticipate the outputs. Surprisingly, after familiarization the perceived utility of the system was rated higher than before, indicating that the perceived utility of AI is not just a novelty effect. The increase in benefits mainly comes from end-users' ability to customize prompts, and thus potentially appropriate the system to their own needs. This points to a future where generative AI systems can allow us to design for appropriation.

Melanie Subbiah, Sean Zhang, Lydia B. Chilton et al.

We evaluate recent Large Language Models (LLMs) on the challenging task of summarizing short stories, which can be lengthy, and include nuanced subtext or scrambled timelines. Importantly, we work directly with authors to ensure that the stories have not been shared online (and therefore are unseen by the models), and to obtain informed evaluations of summary quality using judgments from the authors themselves. Through quantitative and qualitative analysis grounded in narrative theory, we compare GPT-4, Claude-2.1, and LLama-2-70B. We find that all three models make faithfulness mistakes in over 50% of summaries and struggle with specificity and interpretation of difficult subtext. We additionally demonstrate that LLM ratings and other automatic metrics for summary quality do not correlate well with the quality ratings from the writers.

Ziyi Wang, Yuxuan Lu, Wenbo Li et al. · gatech, microsoft-research

Can large language models (LLMs) accurately simulate the next web action of a specific user? While LLMs have shown promising capabilities in generating ``believable'' human behaviors, evaluating their ability to mimic real user behaviors remains an open challenge, largely due to the lack of high-quality, publicly available datasets that capture both the observable actions and the internal reasoning of an actual human user. To address this gap, we introduce OPERA, a novel dataset of Observation, Persona, Rationale, and Action collected from real human participants during online shopping sessions. OPERA is the first public dataset that comprehensively captures: user personas, browser observations, fine-grained web actions, and self-reported just-in-time rationales. We developed both an online questionnaire and a custom browser plugin to gather this dataset with high fidelity. Using OPERA, we establish the first benchmark to evaluate how well current LLMs can predict a specific user's next action and rationale with a given persona and <observation, action, rationale> history. This dataset lays the groundwork for future research into LLM agents that aim to act as personalized digital twins for human.

Tao Long, Xuanming Zhang, Sitong Wang et al.

Agentic workflows promise efficiency, but adoption hinges on whether people actually trust systems that act on their behalf. We present DoubleAgents, an agentic planning tool that embeds transparency and control through user intervention, value-reflecting policies, rich state visualizations, and uncertainty flagging for human coordination tasks. A built-in respondent simulation generates realistic scenarios, allowing users to rehearse, refine policies, and calibrate their reliance before live use. We evaluate DoubleAgents in a two-day lab study (n=10), two deployments (n=2), and a technical evaluation. Results show that participants initially hesitated to delegate but grew more reliant as they experienced transparency, control, and adaptive learning during simulated cases. Deployment results demonstrate DoubleAgents' real-world relevance and usefulness, showing that the effort required scaled appropriately with task complexity and contextual data. We contribute trust-by-design patterns and mechanisms for proactive AI -- consistency, controllability, and explainability -- along with simulation as a safe path to build and calibrate trust over time.

Jenny Ma, Riya Sahni, Karthik Sreedhar et al.

Multi-agent large language model simulations have the potential to model complex human behaviors and interactions. If the mechanics are set up properly, unanticipated and valuable social dynamics can surface. However, it is challenging to consistently enforce simulation mechanics while still allowing for notable and emergent dynamics. We present AgentDynEx, an AI system that helps set up simulations from user-specified mechanics and dynamics. AgentDynEx uses LLMs to guide users through a Configuration Matrix to identify core mechanics and define milestones to track dynamics. It also introduces a method called \textit{nudging}, where the system dynamically reflects on simulation progress and gently intervenes if it begins to deviate from intended outcomes. A technical evaluation found that nudging enables simulations to have more complex mechanics and maintain its notable dynamics compared to simulations without nudging. We discuss the importance of nudging as a technique for balancing mechanics and dynamics of multi-agent simulations.

Tao Long, Dorothy Zhang, Grace Li et al.

Communicating science and technology is essential for the public to understand and engage in a rapidly changing world. Tweetorials are an emerging phenomenon where experts explain STEM topics on social media in creative and engaging ways. However, STEM experts struggle to write an engaging "hook" in the first tweet that captures the reader's attention. We propose methods to use large language models (LLMs) to help users scaffold their process of writing a relatable hook for complex scientific topics. We demonstrate that LLMs can help writers find everyday experiences that are relatable and interesting to the public, avoid jargon, and spark curiosity. Our evaluation shows that the system reduces cognitive load and helps people write better hooks. Lastly, we discuss the importance of interactivity with LLMs to preserve the correctness, effectiveness, and authenticity of the writing.

Yulun Du, Lydia Chilton

Collaborative stories, which are texts created through the collaborative efforts of multiple authors with different writing styles and intentions, pose unique challenges for NLP models. Understanding and generating such stories remains an underexplored area due to the lack of open-domain corpora. To address this, we introduce STORYWARS, a new dataset of over 40,000 collaborative stories written by 9,400 different authors from an online platform. We design 12 task types, comprising 7 understanding and 5 generation task types, on STORYWARS, deriving 101 diverse story-related tasks in total as a multi-task benchmark covering all fully-supervised, few-shot, and zero-shot scenarios. Furthermore, we present our instruction-tuned model, INSTRUCTSTORY, for the story tasks showing that instruction tuning, in addition to achieving superior results in zero-shot and few-shot scenarios, can also obtain the best performance on the fully-supervised tasks in STORYWARS, establishing strong multi-task benchmark performances on STORYWARS.

Katy Ilonka Gero, Lydia B. Chilton, Chris Melancon et al.

Handwriting recognition is improving in leaps and bounds, and this opens up new opportunities for stylus-based interactions. In particular, note-taking applications can become a more intelligent user interface, incorporating new features like autocomplete and integrated search. In this work we ran a gesture elicitation study, asking 21 participants to imagine how they would interact with an imaginary, intelligent note-taking application. We report agreement on the elicited gestures, finding that while existing common interactions are prevalent (like double taps and long presses) a number of more novel interactions (like dragging selected items to hotspots or using annotations) were also well-represented. We discuss the mental models participants drew on when explaining their gestures and what kind of feedback users might need to move to more stylus-centric interactions.

Sitong Wang, Savvas Petridis, Taeahn Kwon et al.

Pop culture is an important aspect of communication. On social media people often post pop culture reference images that connect an event, product or other entity to a pop culture domain. Creating these images is a creative challenge that requires finding a conceptual connection between the users' topic and a pop culture domain. In cognitive theory, this task is called conceptual blending. We present a system called PopBlends that automatically suggests conceptual blends. The system explores three approaches that involve both traditional knowledge extraction methods and large language models. Our annotation study shows that all three methods provide connections with similar accuracy, but with very different characteristics. Our user study shows that people found twice as many blend suggestions as they did without the system, and with half the mental demand. We discuss the advantages of combining large language models with knowledge bases for supporting divergent and convergent thinking.

Katy Ilonka Gero, Chris Kedzie, Savvas Petridis et al.

Language models are known to produce vague and generic outputs. We propose two unsupervised decoding strategies based on either word-frequency or point-wise mutual information to increase the specificity of any model that outputs a probability distribution over its vocabulary at generation time. We test the strategies in a prompt completion task; with human evaluations, we find that both strategies increase the specificity of outputs with only modest decreases in sensibility. We also briefly present a summarization use case, where these strategies can produce more specific summaries.

Katy Ilonka Gero, Vivian Liu, Lydia B. Chilton

Large-scale language models are rapidly improving, performing well on a wide variety of tasks with little to no customization. In this work we investigate how language models can support science writing, a challenging writing task that is both open-ended and highly constrained. We present a system for generating "sparks", sentences related to a scientific concept intended to inspire writers. We find that our sparks are more coherent and diverse than a competitive language model baseline, and approach a human-created gold standard. In a study with 13 PhD students writing on topics of their own selection, we find three main use cases of sparks: aiding with crafting detailed sentences, providing interesting angles to engage readers, and demonstrating common reader perspectives. We also report on the various reasons sparks were considered unhelpful, and discuss how we might improve language models as writing support tools.

Vivian Liu, Lydia B. Chilton

Text-to-image generative models are a new and powerful way to generate visual artwork. However, the open-ended nature of text as interaction is double-edged; while users can input anything and have access to an infinite range of generations, they also must engage in brute-force trial and error with the text prompt when the result quality is poor. We conduct a study exploring what prompt keywords and model hyperparameters can help produce coherent outputs. In particular, we study prompts structured to include subject and style keywords and investigate success and failure modes of these prompts. Our evaluation of 5493 generations over the course of five experiments spans 51 abstract and concrete subjects as well as 51 abstract and figurative styles. From this evaluation, we present design guidelines that can help people produce better outcomes from text-to-image generative models.

Daniel Li, Thomas Chen, Albert Tung et al.

Every day we are surrounded by spoken dialog. This medium delivers rich diverse streams of information auditorily; however, systematically understanding dialog can often be non-trivial. Despite the pervasiveness of spoken dialog, automated speech understanding and quality information extraction remains markedly poor, especially when compared to written prose. Furthermore, compared to understanding text, auditory communication poses many additional challenges such as speaker disfluencies, informal prose styles, and lack of structure. These concerns all demonstrate the need for a distinctly speech tailored interactive system to help users understand and navigate the spoken language domain. While individual automatic speech recognition (ASR) and text summarization methods already exist, they are imperfect technologies; neither consider user purpose and intent nor address spoken language induced complications. Consequently, we design a two stage ASR and text summarization pipeline and propose a set of semantic segmentation and merging algorithms to resolve these speech modeling challenges. Our system enables users to easily browse and navigate content as well as recover from errors in these underlying technologies. Finally, we present an evaluation of the system which highlights user preference for hierarchical summarization as a tool to quickly skim audio and identify content of interest to the user.

Katy Gero, Chris Kedzie, Jonathan Reeve et al.

Despite the success of style transfer in image processing, it has seen limited progress in natural language generation. Part of the problem is that content is not as easily decoupled from style in the text domain. Curiously, in the field of stylometry, content does not figure prominently in practical methods of discriminating stylistic elements, such as authorship and genre. Rather, syntax and function words are the most salient features. Drawing on this work, we model style as a suite of low-level linguistic controls, such as frequency of pronouns, prepositions, and subordinate clause constructions. We train a neural encoder-decoder model to reconstruct reference sentences given only content words and the setting of the controls. We perform style transfer by keeping the content words fixed while adjusting the controls to be indicative of another style. In experiments, we show that the model reliably responds to the linguistic controls and perform both automatic and manual evaluations on style transfer. We find we can fool a style classifier 84% of the time, and that our model produces highly diverse and stylistically distinctive outputs. This work introduces a formal, extendable model of style that can add control to any neural text generation system.

Quanze Chen, Jonathan Bragg, Lydia B. Chilton et al.

Traditional approaches for ensuring high quality crowdwork have failed to achieve high-accuracy on difficult problems. Aggregating redundant answers often fails on the hardest problems when the majority is confused. Argumentation has been shown to be effective in mitigating these drawbacks. However, existing argumentation systems only support limited interactions and show workers general justifications, not context-specific arguments targeted to their reasoning. This paper presents Cicero, a new workflow that improves crowd accuracy on difficult tasks by engaging workers in multi-turn, contextual discussions through real-time, synchronous argumentation. Our experiments show that compared to previous argumentation systems which only improve the average individual worker accuracy by 6.8 percentage points on the Relation Extraction domain, our workflow achieves 16.7 percentage point improvement. Furthermore, previous argumentation approaches don't apply to tasks with many possible answers; in contrast, Cicero works well in these cases, raising accuracy from 66.7% to 98.8% on the Codenames domain.