"Should I choose a smaller model?'': Understanding ML Model Selection and Its Impact on SustainabilityThe increasing accessibility of large machine learning (ML) models has resulted in their widespread adoption in everyday products, with a correspondingly negative environmental impact. Selecting more suitable ML models could not only improve training time and achievable accuracy, but also long-term sustainability. However, ML developers' model selection process remains underexplored, especially with respect to sustainability trade-offs. Our interviews with 13 ML developers showed that participants select models mainly based on familiarity, accuracy and interpretability, but often overlook sustainability. They critically reflected on the current trends of large models and the lack of available information regarding model sustainability. We present implications for the ML and HCI communities, emphasizing the importance of critical reflection on model selection in education and practice. Based on our insights, we provide initial recommendations for promoting model sustainability evaluation and how the HCI community can assist in making sustainable model alternatives more accessible.2025ECEya Ben chaaben et al.Inria Paris Saclay, ExSituAI-Assisted Decision-Making & AutomationSustainable HCIEcological Design & Green ComputingCHI
Interaction Substrates: Combining Power and Simplicity in Interactive SystemsToday’s graphical user interfaces tend to be either simple but limited, or powerful but overly complex. In order to combine power and simplicity, we introduce Substrates, which act as “places for interaction” where users can manipulate objects of interest in a principled and predictable way. Substrates structure and contain data, enforce user-defined constraints among objects and manage dependencies with other substrates. Users can “tune” and “tweak” these relationships, “curry” specialized tools or abstract relationships into interactive templates. We first define substrates and provide in-depth descriptions with examples of their key characteristics. After explaining how Substrates extend the concept of Instrumental Interaction, we apply a Generative Theory of Interaction approach to analyze and critique existing interfaces and then show how using the concepts of Instruments and Substrates inspired novel design ideas in three graduate-level HCI courses. We conclude with a discussion and directions for future work.2025WMWendy E Mackay et al.Inria, ExSituPrototyping & User TestingCHI
DesignPrompt: Using Multimodal Interaction for Design Exploration with Generative AIVisually oriented designers often struggle to create effective generative AI (GenAI) prompts. A preliminary study identified specific issues in composing and fine-tuning prompts, as well as needs in accurately translating intentions into rich input. We developed DesignPrompt, a moodboard tool that lets designers combine multiple modalities — images, color, text — into a single GenAI prompt and tweak the results. We ran a comparative structured observation study with 12 professional designers to better understand their intent expression, expectation alignment and transparency perception using DesignPrompt and text input GenAI. We found that multimodal prompt input encouraged designers to explore and express themselves more effectively. Designer’s interaction preferences change according to their overall sense of control over the GenAI and whether they are seeking inspiration or a specific image. Designers developed innovative uses of DesignPrompt, including developing elaborate multimodal prompts and creating a multimodal prompt pattern to maximize novelty while ensuring consistency.2024XPXiaohan Peng et al.Generative AI (Text, Image, Music, Video)Graphic Design & Typography ToolsCreative Collaboration & Feedback SystemsDIS
What Counts as ‘Creative’ Work? Articulating Four Epistemic Positions in Creativity-Oriented HCI ResearchThis paper examines prevailing understandings of creativity within creative computing research through the lens of feminist epistemology. We analyze creativity support as a construct that encodes different definitions of creative work. Drawing on existing literature and practices, the paper surfaces four views about creative work that underpin current creative technologies and HCI research: problem-solving, cognitive emergence, embodied action, and tool-mediate expert activity. Each view makes different claims about the role of computing in creative work and the creative subject assumed. We articulate the attendant politics of each view and illustrate how critical feminist epistemology can serve as an analytical tool to reason about the trade-offs of various creativity definitions. The paper concludes with suggestions on integrating feminist values into creativity-oriented HCI research.2024SHStacy Hsueh et al.KTH Royal Institute of TechnologyGenerative AI (Text, Image, Music, Video)Technology Ethics & Critical HCICHI
Mirrorverse: Live Tailoring of Video Conferencing InterfacesHow can we let users adapt video-based meetings as easily as they rearrange furniture in a physical meeting room? We describe a design space for video conferencing systems that includes a five-step ``ladder of tailorability,'' from minor adjustments to live reprogramming of the interface. We then present Mirrorverse and show how it applies the principles of computational media to support live tailoring of video conferencing interfaces to accommodate highly diverse meeting situations. We present multiple use scenarios, including a virtual workshop, an online yoga class, and a stand-up team meeting to evaluate the approach and demonstrate its potential for new, remote meetings with fluid transitions across activities.2023JGJens Emil Sloth Grønbæk et al.Remote Work Tools & ExperienceUIST
Readymades & Repertoires: Artifact-Mediated Improvisation in Tabletop Role-Playing GamesGame masters (GMs) are creative practitioners who plan and orchestrate tabletop role-playing games. Through an interview study, we investigate how eight expert game masters adapt everyday technologies and materials as creativity support tools \added{(CSTs)} for improvisational and collaborative play. We integrate theories of improvisational and distributed creativity with the human-artifact model, which provides an activity-theoretical vocabulary for analyzing the mediating relationships between specialist practitioners and their tools. We show how GMs prepare and deploy readymade artifacts: analog and digital CSTs that flexibly mediate recurring creative tasks in their practice, such as improvising narrative elements, facilitating smooth play, and creating aesthetic effects. We find that GMs demonstrate designerly thinking as they create, share, and refine repertoires of readymade artifacts. We argue that our theoretical approach can inform future studies of IT-mediated creativity, and that readymade artifacts can be an analytical and generative concept for the design of novel creativity support tools.2022PTPhilip Tchernavskij et al.Role-Playing & Narrative GamesInteractive Narrative & Immersive StorytellingC&C
Deep Learning Uncertainty in Machine TeachingMachine Learning models can output confident but incorrect predictions. To address this, ML researchers use various techniques to reliably estimate ML uncertainty, usually performed on controlled benchmarks once the model has been trained. We explore how the two types of uncertainty—aleatoric and epistemic—can help non-expert users understand the strengths and weaknesses of a classifier in an interactive setting. We are interested not only in their use of uncertainty to teach and understand the classifier, but also in their perception of the difference between aleatoric and epistemic uncertainty. We conducted an experiment where non-experts train a classifier to recognize card images, and are tested on their ability to predict classifier outcomes. Participants who used either larger or more varied training sets significantly improved their understanding of uncertainty, both epistemic or aleatoric. However, participants who relied on the uncertainty measure to guide their choice of training data did not significantly improve classifier training, nor were they better able to guess the classifier outcome. We identified three specific situations where participants successfully identified the difference between aleatoric and epistemic uncertainty: placing a new card in the exact same position as a training card; placing different cards next to each other; and placing a non-card, such as their hand, next to or on top of a card. We discuss our methodology for estimating uncertainty for Interactive Machine Learning systems and question the need for two-level uncertainty in Machine Teaching.2022TSTéo Sanchez et al.Explainable AI (XAI)AI-Assisted Decision-Making & AutomationIUI
Dance Transitions: What Forms of Technology Best Support Professional Dancers as They Learn New Movement Styles?Most current dance support technologies focus on dancers, teachers or choreographers who are engaged in a single activity. We are interested in creating tools that support professional dancers over longer periods of time, as their careers and personal practices evolve. We interviewed 12 professional and pre-professional dancers about a critical moment in their careers: the transition to a new dance style due to shifting interests, ageing or injury. We identify three key challenges -- overcoming habits, learning new forms of movement, transitioning over time -- and their strategies for addressing them. We argue that successful tools must help dancers change their mentality about new movement styles, rather than focusing solely on movement mechanics. We suggest three possible implications for design: develop ``movement substrates'' that handle multiple movement representations; integrate learning and reflection in a single session; and create movement definitions through movement. We conclude with a discussion of directions for future research.2022EWElizabeth Walton et al.CNRS, InriaFull-Body Interaction & Embodied InputFoot & Wrist InteractionDance & Body Movement ComputingCHI
Exploring the Role of Artifacts in Collective Dance Re-stagingPreparing a new dance performance involves more than asking dancers to learn individual steps. We are interested in understanding how dancers collaborate as they rehearse a new dance piece, with a particular emphasis on how they use physical and digital artifacts to support this process. We conducted a 12-month longitudinal observation study with a dance company that re-staged a dance piece, taken from the contemporary repertoire and unknown to the dancers. This study focuses on the role that artifacts, used during the rehearsal process, play in shaping learning. We show how dancers produced an ecology of artifacts with the aim of distributing their knowledge and sharing it with other learners. We show how artifacts serve to decompose the choreography into simpler components, independent and complementary, with the objective to reduce the difficulty of the learning task. We found that dancers compile artifacts to create a common structure among the group, allowing for improving the learning process. We conclude with design opportunities for technologies supporting long-term dance learning processes.2021JRJean-Philippe Riviere et al.Specialist and Collaborative Work // Algorithmic FairnessCSCW
SonAmi: A Tangible Creativity Support Tool for Productive ProcrastinationOur goal is to help creative writers make procrastination productive. Interviews with eight creative writers highlighted two key practices: Over Criticizing, where perfectionism and negative self-appraisal demotivates them and reduces their output; and Creative Voicing, where speaking their text aloud promotes reflection and inspires new possibilities. A structured observation study compared writers’ perception of their own, pre-recorded text versus computer-generated voices. The latter distances them from their text and offers new perspectives. We developed SonAmi, an interactive coaster that voices selected dialog whenever the author lifts their mug. Two creative writers said SonAmi made them feel they were listening to “someone else’s text” or “a podcast”, which helped them identify and improve writing issues. We show how tangible creativity support tools can build upon authors’ existing strategies i.e., voicing their own words, and take advantage of naturally occurring events e.g., taking a sip of coffee, to support productive procrastination without interfering with creative flow.2021JBJekaterina Belakova et al.Force Feedback & Pseudo-Haptic WeightAI-Assisted Creative WritingC&C
Textlets: Supporting Constraints and Consistency in Text DocumentsWriting technical documents frequently requires following constraints and consistently using domain-specific terms. We interviewed 12 legal professionals and found that they all use a standard word processor, but must rely on their memory to manage dependencies and maintain consistent vocabulary within their documents. We introduce Textlets, interactive objects that reify text selections into persistent items. We show how Textlets help manage consistency and constraints within the document, including selective search and replace, word count, and alternative wording. Eight participants tested a search-and-replace Textlet as a technology probe. All successfully interacted directly with the Textlet to perform advanced tasks; and most (6/8) spontaneously generated a novel replace-all-then-correct strategy. Participants suggested additional ideas, such as supporting collaborative editing over time by embedding a Textlet into the document to flag forbidden words. We argue that Textlets serve as a generative concept for creating powerful new tools for document editing.2020HHHan L. Han et al.Université Paris-Saclay, CNRS, InriaKnowledge Worker Tools & WorkflowsPrototyping & User TestingCHI
Customizations and Expression Breakdowns in Ecosystems of Communication AppsThe growing adoption of emojis, stickers and GIFs suggests a corresponding demand for rich, personalized expression in messaging apps. Some people customize apps to enable more personal forms of expression, yet we know little about how such customizations shape everyday communication. Since people increasingly communicate via multiple apps side-by-side, we are also interested in how customizing one app influences communication via other apps. We created a taxonomy of customization options based on interviews with 15 "extreme users" of communication apps. We found that participants tailored their apps to express their identities, organizational culture, and intimate bonds with others. They also experienced expression breakdowns: frustrations around barriers to transferring personal forms of expression across apps, which inspired inventive workarounds to maintain cross-app habits of expression, such as briefly switching apps to generate and export content for a particular conversation. We conclude with implications for personalized expression in ecosystems of communication apps.2019CGCarla Griggio et al.Language and Expressivity ICSCW
Augmenting Couples' Communication with Lifelines: Shared Timelines of Mixed Contextual InformationCouples exhibit special communication practices, but apps rarely offer couple-specific functionality. Research shows that sharing streams of contextual information (e.g. location, motion) helps couples coordinate and feel more connected. Most studies explored a single, ephemeral stream; we study how couples' communication changes when sharing multiple, persistent streams. We designed Lifelines, a mobile-app technology probe that visualizes up to six streams on a shared timeline: closeness to home, battery level, steps, media playing, texts and calls. A month-long study with nine couples showed that partners interpreted information mostly from individual streams, but also combined them for more nuanced interpretations. Persistent streams allowed missing data to become meaningful and provided new ways of understanding each other. Unexpected patterns from any stream can trigger calls and texts, whereas seeing expected data can replace direct communication, which may improve or disrupt established communication practices. We conclude with design implications for mediating awareness within couples.2019CGCarla F. Griggio et al.University Paris-Sud, CNRS, Inria & Université Paris-SaclaySocial Platform Design & User BehaviorContext-Aware ComputingUbiquitous ComputingCHI
Touchstone2: An Interactive Environment for Exploring Trade-offs in HCI Experiment DesignTouchstone2 offers a direct-manipulation interface for generating and examining trade-offs in experiment designs. Based on interviews with experienced researchers, we developed an interactive environment for manipulating experiment design parameters, revealing patterns in trial tables, and estimating and comparing statistical power. We also developed TSL, a declarative language that precisely represents experiment designs. In two studies, experienced HCI researchers successfully used Touchstone2 to evaluate design trade-offs and calculate how many participants are required for particular effect sizes. We discuss Touchstone2's benefits and limitations, as well as directions for future research.2019AEAlexander Eiselmayer et al.University of ZurichUser Research Methods (Interviews, Surveys, Observation)Prototyping & User TestingComputational Methods in HCICHI
Interstices: Sustained Spatial Relationships between Hands and Surfaces Reveal Anticipated ActionOur observations of landscape architecture students revealed a new phenomenon—interstices. Their bimanual interactions with a pen and touch surface involved various sustained hand gestures, interleaved between their regular commands. Positioning of the non-preferred hand indicates anticipated actions, including: sustained hovering near the surface; pulled back but still floating above the surface; and resting in their laps. We ran a second study with 14 landscape architect students which confirmed our observations, and uncovered a new interstice i.e. stabilizing the preferred hand while handwriting. We conclude with directions for future research and challenges for designers and researchers.2019AWAndrew M. Webb et al.Inria, Université Paris-Saclay and Texas A&M UniversityFull-Body Interaction & Embodied InputPrototyping & User TestingCHI
Rethinking Interaction: From Instrumental Interaction to Human-Computer PartnershipsThe extraordinary advances in hardware and networking technology over the past 50 years have not been matched by equivalent advances in software. Today’s interactive systems are fraught with limitations and incompatibilities: they lack interoperability and flexibility for end users. The goal of this workshop is to rethink interaction by identifying frameworks, principles and approaches that break these limitations and create true human-computer partnerships.2018MBMichel Beaudouin-Lafon et al.Univ. Paris-Sud, CNRS, Inria, Université Paris-SaclayHuman-LLM CollaborationAI-Assisted Decision-Making & AutomationCHI
BIGFile: Bayesian Information Gain for Fast File RetrievalWe introduce BIGFile, a new fast file retrieval technique based on the Bayesian Information Gain framework. BIGFile provides interface shortcuts to assist the user in navigating to a desired target (file or folder). BIGFile’s split interface combines a traditional list view with an adaptive area that displays shortcuts to the set of file paths estimated by our computationally efficient algorithm. Users can navigate the list as usual, or select any part of the paths in the adaptive area. A pilot study of 15 users informed the design of BIGFile, revealing the size and structure of their file systems and their file retrieval practices. Our simulations show that BIGFile outperforms Fitchett et al.’s AccessRank, a best-of-breed prediction algorithm. We conducted an experiment to compare BIGFile with ARFile (AccessRank instantiated in a split interface) and with a Finder-like list view as baseline. BIGFile was by far the most efficient technique (up to 44% faster than ARFile and 64% faster than Finder), and participants unanimously preferred the split interfaces to the Finder.2018WLWanyu Liu et al.Telecom ParisTech, Université Paris-Saclay, Univ. Paris-Sud, CNRS, Inria, Université Paris-SaclayInteractive Data VisualizationTime-Series & Network Graph VisualizationVisualization Perception & CognitionCHI
Knotation: Exploring and Documenting Choreographic ProcessesContemporary choreographers often interact directly with dancers when exploring their ideas, but lack adequate tools for capturing and documenting their work. Although our first study of choreographers and dancers revealed diverse strategies for recording choreographic fragments, we found that they all worked in terms of constraints, which they represented via spatial diagrams, as movement qualities or with their own personal notation system. This led to the design of Knotation, a mobile pen-based tool that lets choreographers sketch their own representations of choreographic ideas and render them interactive. In study two, Knotation served as a technology probe to support the contrasting practices of three professional choreographers. We revised Knotation based on their input, and ran a third structured observation study with six professional choreographers. Knotation easily supported both dance-then-record and record-then-dance strategies. Participants used and appropriated Knotation's advanced features, including the combination of interactive timelines and floorplan diagrams, to represent and explore complex choreographic structures.2018MFMarianela Ciolfi Felice et al.Univ. Paris-Sud, CNRS, Inria, Université Paris-SaclayDance & Body Movement ComputingCHI