ICIDS 2026: INTERNATIONAL CONFERENCE ON INTERACTIVE DIGITAL STORYTELLING 2026
PROGRAM FOR TUESDAY, DECEMBER 1ST
Days:
previous day
next day
all days

View: session overviewtalk overview

10:30-10:45Coffee Break
10:45-12:45 Session 7: Paper Session: Generative AI Part 1
10:45
When AI Estimates and Humans Decide: Fostering Understanding of AI-Assisted Medical Decision Making Through Interactive Narrative

ABSTRACT. AI-assisted systems are increasingly used to support high-stakes decisions, yet the processes by which AI-generated predictions inform human judgment often remain opaque to learners. As adolescents grow up in an AI-enabled world, they need opportunities to develop AI literacy that extends beyond technical concepts to understanding how AI-generated outputs interact with human judgment, institutional policies, and professional decision-making processes. Interactive narratives offer a promising approach by embedding learning within investigative experiences. This paper presents an episode of a narrative-centered learning environment designed to foster adolescents’ understanding of AI-assisted medical decision-making. In the interactive narrative, learners assume the role of a medical intern investigating why patients with similar symptoms receive different triage outcomes. Using patient cases, healthcare conversations, and AI-generated risk estimates, learners reason about how human oversight and organizational policies shape medical decisions. We report findings from a mixed-methods pilot study involving seven adolescents who completed the interactive narrative and participated in a focus group discussion. Data sources included pre- and post-intervention AI knowledge assessments, self-report measures, and qualitative focus group data. Findings suggest that participants developed a more nuanced understanding of how AI-generated risk estimates inform clinical decision making and recognized that healthcare professionals remain responsible for interpreting those estimates and determining action. Participants also experienced challenges interpreting risk estimates and understanding how threshold-based policies influence outcomes. We discuss implications for designing interactive narratives that make complex human-AI decision processes accessible to young learners and highlight the potential of investigative narrative structures for supporting AI literacy in sociotechnical contexts.

11:15
AI as Co-coder to Address the Data Analysis Challenge in Interactive Digital Narrative Research

ABSTRACT. The capacity for modern AI to process language offers opportunities to study dynamic interactive digital narrative (IDN) system-process-product interactions and their impacts. Nevertheless, because IDN involves a complex interweaving of multiple expressive media, linear cause-and-effect research paradigms that assume a singular voice may not be adequate. IDN researchers are now incorporating AI agents as interlocutors, further increasing complexity. We argue for com-plex in-process IDN research methods, presenting a case study evaluating hu-man-AI collaboration that alleviates data analysis burden through an interactive co-coding process. With diverse reliability analyses within and between an expert data analyst and an LLM agent, this inquiry points to the need for a purposeful mediator: a human-designed Codebook. Based on iterative analytic prompting procedures and findings, the Codebook proved an effective, reproducible model for managing co-coding of diverse language genres in and around IDN. Results indicated that the LLM's value was diagnosing the Codebook, as it behaved differently on two IDN language dimensions: narrative and procedural elements versus player-designer conversational pragmatics. Rather than replicating the hu-man data analysis expert, the LLM thus served as a dimensional dial calibrating diverse language forms and processes, and revealed diverse human and AI capacities and limits. Contributions of this study include offering a model for human-AI practice and research to address what promises to be an increasing data analysis challenge in IDN. At stake is the future of scientific inquiry in IDN and its impact on human development, learning, well-being, and social change.

11:45
LLM-Powered Characters as Narrative Companions: An Interaction Analysis of Character.AI Chatbots
PRESENTER: Kexin Xiang

ABSTRACT. This paper aims to theorize roleplaying practices on AI companion plat-forms and similar LLM-based applications as an emerging form of interac-tive digital narrative (IDN). Focusing on Character.AI’s one-on-one roleplay mode, we conceptualize AI companion roleplay as “LLM-powered single-player online freeform roleplay” and analyze its distinctive narrative opera-tions. Through app walkthrough and interaction analysis of recorded roleplay sessions, we develop a conceptual toolkit for understanding this IDN form, including “fictional play space,” defined as the range of possible storyworlds afforded by an IDN system, and four recurring frames of player-AI narrative interaction: “creative,” “elicitive,” “elaborative,” and “regres-sive.” These frames support a gravity model of Character.AI’s fictional play spaces, in which player-driven storyworld expansion continually encoun-ters model-maintained defaults. We further argue that the appeal of AI com-panion roleplay depends on a “companion principle”: the extent to which an AI character can effectively sustain and inspire players’ creative expan-sion of the storyworld. Together, these concepts suggest “narrative compan-ionship” as an emerging mode of IDN engagement in the age of AI, where individual players gain greater narrative agency while encountering new constraints, and in which AI characters are expected to function as respon-sive narrative companions despite the complex design challenges involved in realizing this role.

12:15
Story Exchange Engines (SEEs) for Decision Support: A Model for AI-Augmented Storytelling in Organisational Decision-Making

ABSTRACT. Organisations already use storytelling in their decision procedures: the pre-mortem, scenario planning, appreciative inquiry, scenario-based design, sensemaking. These procedures were limited not by theory but by the practical cost of running stories at scale. AI lifts that ceiling. We ask what common structure underlies this use of story, and how AI changes it. We examine three cases: interaction design at a museum (Moyland), participatory urban planning (Duisburg), and the pre-mortem (Military). From them we derive a model: a narrative exchange between a reference story and counter-stories, with a return that revises the reference; nine dimensions along which exchanges vary; tensions between the dimensions; and a Story Exchange Engine (SEE) with eight faculties, each paired with a shadow. The engine assists, manages, and keeps form and norms; authoring, issuing, and deciding remain with the participants. A gradual stake criterion bounds the delegation: the more an act has at stake, the less delegable it becomes. The method is case-grounded design fiction: the cases ground the model; the fictions generate testable research questions.

12:45-13:45Lunch Break
13:45-15:15 Session 8: Paper Session: Interactive Storytelling for Reflection and Therapy Part 1
13:45
Creating Interactive Stories for Reflection: How College Students Author Nonlinear Narratives about Stressful Experiences

ABSTRACT. Interactive digital narratives (IDNs) are nonlinear stories where the user's decisions have an impact on the narrative's outcomes. Prior literature proposes that the reflection process needed for the creation of IDNs may have benefits for one's mental wellbeing. Existing tools for IDN authoring are focused on supporting the creation of nonlinear stories, and do not include scaffolds to facilitate the reflection process needed for nonlinear story creation to support mental wellbeing. To understand how IDN authoring tools can be designed to foster reflection, we conducted a study with college students using an IDN authoring tool to create a non-linear story about a recent stressful experience. Findings from participants' story creation process and retrospective interviews include themes on how they approached the task of creating a nonlinear story. From these, we contribute an initial process model of nonlinear authoring reflection, and present design implications for scaffolds needed in IDN authoring tools to support independent, reflective authoring.

14:15
Towards a Generative Design Space for Reflective Questions in Interactive Narrative Experiences
PRESENTER: Maria Vayanou

ABSTRACT. Reflective questions are widely used in cultural heritage, educational, and interactive narrative experiences to support interpretation, perspective taking, empathy, and meaning making. Despite their prevalence, little attention has been devoted to understanding how reflective questions themselves are designed. This paper proposes a generative design space for reflective questions based on an analysis of 122 reflective prompts drawn from six cultural heritage and museum experiences. Through an iterative, theory-informed analysis, we identify three complementary design dimensions: Reflection Targets (what participants reflect upon), Reflective Focus (how reflection is framed through affective, behavioural, or cognitive engagement), and Perspective Frames (the person, time, and place from which reflection is performed). We demonstrate how these dimensions can be used both to generate alternative reflective questions and to analyse existing experiences through concepts such as reflective profiles, scaffolding strategies, and perspective trajectories. Beyond providing a vocabulary for reflective design, the framework offers opportunities for AI-assisted authoring and future empirical studies investigating the effects of alternative reflective strategies on learning, empathy, and transformative reflection.

14:45
Designing Safe Antagonism: Enhancing a Therapeutic Narrative Board Game with a Digital Companion
PRESENTER: Maria Goikhman

ABSTRACT. This paper presents the redesign of Mind Shadows, a therapeutic storytelling board game in which players collaboratively engage with an antagonistic force known as the Shadow. Research-through-design, internal play-testing and user evaluation were used to identify issues related to role conflict, rule clarity, accessibility, and the management of the Shadow. The redesign explores the concept of safe antagonism through a digital companion tool that externalizes the antagonist as a separate presence and offloads its management, while preserving player-authored content. Initial evaluation suggests that the companion app enhances collaboration while improving usability and visual clarity, while rule clarity and intuitiveness remained key challenges. The contribution of this work lies in its exploration of how digital companions can enrich board games by supporting complex antagonistic narrative roles in therapeutic play contexts.

15:15-15:30Coffee Break
15:30-17:00 Session 9: Paper Session: Authoring
15:30
Aleatory Drama Management: How Anchor Threads Answer an Old Problem of Emergent Narrative

ABSTRACT. Emergent narrative research has long treated the tension between bottom-up emergence and top-down story management as an open design problem. This paper describes a vernacular storytelling tradition that has been evolving a working settlement for two decades: Japanese anchor threads (anka threads and their dice-driven descendants, anko threads), in which a single author improvises serialized fiction on anonymous bulletin boards while delegating narrative decisions to reader posts and public dice rolls. Drawing on community self-documentation and a pilot corpus of nine works (2016–2024, eight tripcode authors, ca. 4,900 authorial posts and 2,300 die rolls), we formalize the form's decision structure and name its governing principle aleatory drama management: a regime in which, we argue, a human drama manager's interventions remain communally acceptable because binding portions of selection authority are visibly surrendered to chance and to the crowd. We trace the form's mechanism evolution as a progressive redistribution of selection authority toward publicly witnessed, ungameable chance, driven by community governance failures, and argue that this vernacular settlement, documented here as a tendency within a small single-culture corpus, reframes the emergent-narrative binary as a problem of legitimacy rather than of algorithm design.

15:45
Interactive Storytelling with Parallel and Conditional Graphs: a Conceptual Model and Visual Representation

ABSTRACT. A variety of approaches exist for designing and modelling interactive narrative content. One critical factor is how easily an author can write this content. We identified a gap between sophisticated, interactive approaches that are too abstract and complex for authors and simple branching or conditional branching approaches. Therefore, we propose an entirely visual model that is more powerful than branching narratives.

The model is presented as a graph comprising various node and link types. It enables to represent conditional branching, probabilistic branching, sub-graphs superposition and counter mechanism. We have applied the model to two use cases, taken from large projects using complex rule-based engines. This tends to show that an authoring tool based on the proposed model may be relevant for interactive narrative authoring.

16:00
Creative Drift in Co-Creative Storytelling: How Prompting, Ambiguity, and Memory Shape Narrative Trajectories

ABSTRACT. Large language models enable new forms of human--AI co-creation in which stories emerge through interactive collaboration between users and generative systems. Unlike traditional text generation, co-creative storytelling unfolds over many conversational turns, producing evolving narrative trajectories rather than a single static output. Although semantic change is a natural part of collaborative storytelling, extended interaction can gradually shift a narrative away from its original goals, themes, or constraints.

We introduce creative drift, a phenomenon describing systematic changes in narrative trajectories during long-horizon human--AI storytelling. Rather than viewing drift solely as inconsistency, we investigate how different interaction conditions influence the stability and evolution of collaborative narratives. We construct a controlled storytelling environment in which a large language model collaborates with a simulated user over multiple turns and evaluate three factors commonly encountered in co-creative interaction: prompt perturbations, ambiguous user inputs, and persistent narrative memory. Experimental results show that minor prompt variations produce measurable differences in narrative development, ambiguous interactions substantially increase narrative divergence, and maintaining an explicit narrative memory significantly reduces long-horizon drift while preserving local narrative coherence. These findings provide empirical evidence that interaction design plays a critical role in maintaining coherent long-term collaboration and offer practical guidance for building more reliable human--AI co-creative systems.

16:30
Hands as Narrators: Characterizing Idea Reuptake from Hand Enactment in Story Writing
PRESENTER: Ila Adhikari

ABSTRACT. Embodied externalization is a productive resource for narrative authoring, yet we know surprisingly little about the return trip: the cognitive process by which storytellers later retrieve the externalized ideas and fold them back into a written text. Building on distributed cognition and embodied story authoring, we call this step idea reuptake. To better understand the characteristics of idea reuptake in embodied story authoring, we conducted a two-part study with 24 university participants who recorded short, hand-enacted stories and, a week later, translated those recordings into text using a custom web platform that tracked gaze patterns and video-scrubbing behavior. Reuptake proved frequent and time-consuming, accounting for roughly a quarter of total writing time. Storytellers adapted their returns to the needs of the moment: we identify four reuptake strategies (Holistic, Recurrent, Concurrent, Specific), five reasons for revisiting the recording (forgetting details, improving quality, recovering from distraction, confirming completion, organizing connections), and a recurrent set of challenges centered on locating specific moments in time-based externalized media. We then translate the documented challenges into concrete design implications for embodied interactive story authoring tools.

17:00-17:15Coffee Break
17:45-18:45 Session 11: Late-Breaking Works Part 1
17:45
Composing Consequence: Sound as an Operative Narrative System in Multi-Participant Interactive Digital Narrative

ABSTRACT. This paper examines how interactive digital narrative can use information asymmetry and distributed agency as narrative forces, and how sound can make their consequences perceptible. It presents [Anonymized Work], a multi-participant interactive narrative in which five participants occupy unequal institutional roles and act on different information. We propose Sound-Operative Narrative Systems (SONS), a framework for systems in which sound participates directly in the production of narrative information and consequence. Its central mechanism, the Role-Conditioned Sonic Attribution Loop (RCSAL), links private sonic evidence, participant action, and a changing shared sound field. Decisions made by one participant thereby alter the conditions through which later participants interpret and act. Through this recursive structure, the work uses sound to carry distributed causal effects across participants and to construct a run- specific history of shared responsibility. The paper describes the current system and outlines its planned evaluation during an October 2026 public presentation.

17:51
Interactive Structured Discussion Forums: Formalizing Structured Forums in Visual Novels through the SPP Model

ABSTRACT. Fostering ethical awareness requires critical thinking beyond binary judgments. In this context, one active learning strategy is to use a visual novel to unfold a case that requires decisions and actions with ethical implications. However, beyond solving cases, ethical learning also rests on the practice of debating. To reconcile the guided nature of VNs with the dialogic engagement needed for ethical reflection, a challenge we address by developing an open learning environment based on Visual Novel (VN) design. To reconcile the guided nature of VNs with the collaborative agency needed for ethical reflection, we introduce Interactive Structured Discussion Forums (ISDF). This architectural framework embeds structured discussion forums directly within the narrative’s decision-making nodes. This approach transforms a typically solitary experience into a collective ethical debate. In this paper, we present an exploratory pilot study that motivated this framework. We then formalize it in the form of an expansion of the System-Process-Product (SPP) model (Koenitz, 2023) to dynamically orchestrate multi-interactor debate within the narrative.

17:57
Is a “Long Playthrough” Still a Replay? Examining Replay in IDNs through a Close Reading of Expelled!

ABSTRACT. There has been a recent trend in story-based games or interactive digital nar-ratives (IDNs) to focus on the need to replay to fully understand the story and complete the game, in what has been referred to as a “long playthrough”. Examples of this type of game include Save the Date, Zero Escape: Zero Time Dilemma, and Elsinore. In these games, rather than replaying from a “clean” starting point, players are brought back to the game’s start, but game state from the previous playthrough persists, and is needed for the player to reach a “true” ending. In addition, the player’s character is often shown to be diegetically aware of this repetition. In this paper, I explore whether this type of replay can really be considered replay, given the fact that, from the perspective of both the player and the player’s character, the previous playthrough has continued. I use this to re-examine Mitchell’s theory of replay and suggest that we need to rethink what we mean by replay in IDNs.

18:03
Route-Based Nonlinearity in Fixed-Truth Detective Games: A Structural and Observational Study of The Painscreek Killings

ABSTRACT. This paper introduces an operational framework---architectural dependency analysis---for distinguishing structurally required location ordering from enacted discovery-order variability in fixed-truth detective games, and proposes route-based nonlinearity as an analytical concept for this distinction. Applied to The Painscreek Killings (EQ Studios, 2017), the analysis identifies five pairwise-incomparable locations (flexible zone) and dependency-constrained locations forming a partial order (structural skeleton). Three playthroughs show observable flexible-zone divergence (mean 5.33 pairwise inversions out of 10) and skeleton-consistent ordering. Different routes yield distinct location-visit trajectories over the same content without altering narrative outcomes. We identify the visit-sequence contexts these routes constitute, not their interpretive effects.

18:09
Playful and Reflective: Visual Metaphor and Game Design for Accessible Autobiographical Interactive Digital Narratives

ABSTRACT. Autobiographical interactive digital narratives tend to be emotionally taxing. Players are often placed in a complicated world with a dark story and difficult consequences, restricted by content warnings, which can make these narratives hard to approach for audiences who might otherwise be interested. This practice-based research asks whether visual metaphor and playful game design can make a personal story more accessible without affecting the meaning. It examines Capicú (2026), an autobiographical sandbox-platformer videogame about growing up in Venezuela and being forced to migrate, exploring how a colourful, cartoony, figured world can carry lived-experience commentary while inviting reflection. Drawing on the concept of procedural rhetoric, visual metaphor, and the Latin American tradition of testimonio, the work reports on public playtesting and asks how these techniques communicate an author's lived experience in ways that are both playful and reflective. We demonstrate this by testing how abstraction and metaphor let players feel connected to a place they had never been, and that a lighter register can broaden who is willing to sit with a difficult story.

18:15
Playing Under the Dome: Exploring Collective Immersion in Planetarium Games

ABSTRACT. Interactive fulldome environments provide an opportunity to explore im-mersive experiences from a different perspective. This paper presents three interactive games developed for the ARTIS Planetarium in collaboration with (anonymous) University as part of the broader Shared Realities re-search program. The games employ different forms of interaction: competi-tive gameplay in Frog Racer, environmental interaction in an underwater experience, and audience-generated content in The World as a Canvas. Drawing on literature on fulldome media, we use these cases to explore how different configurations of interaction and spectatorship may contribute to immersion. The paper proposes a research agenda combining structured au-dience observation, interviews, and iterative design experiments to investi-gate how immersion is produced, distributed, and shared among differently participating members of a fulldome audience.

18:21
Branching through the Archive: Intent-Aware, LLM-Grounded Interactive Narrative Discovery for Historical Digital Collections
PRESENTER: Yi-Shuai Xu

ABSTRACT. Historical digital collections are increasingly accessible online, yet their interfaces often remain centered on keyword search and ranked result lists, offering limited support for open-ended narrative exploration. This Late Breaking Work investigates how large language models can help transform retrieved collection records into interactive narrative pathways for cultural-heritage discovery. We present an intent-decomposed, metadata-grounded pipeline that combines semantic retrieval, balanced evidence construction, constrained narrative generation, and interactive branching. The approach was implemented using 21 historical Bangkok photographs from the Library of Congress World’s Transportation Commission Photograph Collection. Multidimensional queries were automatically decomposed into complementary retrieval intents, from which balanced evidence pools were constructed. Gemini 3.1 Flash Lite then generated short metadata-grounded narratives and three evidence-linked branches that allow users to continue exploration along alternative thematic directions. A preliminary evaluation examined retrieval behaviour, multi-intent coverage, generation-constraint compliance, claim-level groundedness, and branch quality. Across ten exploratory queries, the mean diagnostic Top-5 thematic relevance rate was 90.0%. In an illustrative multidimensional scenario, intent decomposition increased dimension coverage from 0.50 to 1.00. All four narrative scenarios satisfied the specified structural generation constraints, while 80.0% of 30 assessed claims were fully supported and 96.7% were at least partially supported by the supplied metadata. The prototype demonstrates the feasibility of reorganizing collection retrieval into branching narrative interaction, while highlighting the importance of distinguishing representational balance, structural evidence linkage, and semantic grounding in LLM-supported historical collection exploration.

18:27
Countering Misinformation During Exposure: An Ongoing Research Program

ABSTRACT. This Late Breaking Report outlines an ongoing research agenda investigating whether misinformation can be countered during exposure to digital content. The project focuses on misinformation presented in narrative and expository form and examines whether a during-exposure intervention influences users' responses to misinformation and whether its effects differ by narrative structure. Study 1 investigates an intervention delivered during exposure, in which corrective information is delivered in text form when users first encounter a misinformation claim. The study examines whether the intervention reduces perceived accuracy of misinformation and whether its effects differ between narrative and expository misinformation. It also investigates transportation and perceived processing interruption as factors associated with responses to the intervention. Study 2 extends this approach by examining whether delivering the correction through a virtual agent (VA), rather than through conventional text-based presentation, influences responses to during-exposure misinformation correction. Together, the studies investigate during-exposure misinformation correction while examining how narrative structure and socially mediated delivery influence users' responses during social media content consumption.

18:33
When the Loop Closes: Withdrawing Ability as Expressive Content in an Interactive Narrative of Erasure

ABSTRACT. This Late-Breaking Work presents The Greenbrook Institute, a completed, playable text-based interactive digital narrative (IDN) in which a teacher is administratively and surgically erased inside a 1950s psychiatric institution. The work introduces a design principle we call co-degradation. Much of the field currently pursues systems that enlarge what the interactor is able to do and to perceive, an ambition intensified by generative AI. Co-degradation runs in the opposite direction: the interactor’s agency and the experience’s expressive bandwidth are placed in a coordinated, structured decline, made isomorphic with the protagonist’s regression, so that the interactor undergoes rather than observes the stripping away of a person’s abilities, stratum by stratum. In most modalities a designer would degrade two separate channels, one governing what the interactor can do (choice) and one governing what the interactor can perceive (expression). In the textual medium the two collapse into a single material, and one act of degradation withdraws meaning and agency together. Read against this conference’s theme, the work amounts to an inquiry into ability conducted by subtraction: the capacities to adapt, to choose, and to mean are treated here as designable resources, resources a work can also be built to take away. We situate the principle against earlier expressive denials of agency and reflect on its stakes.

18:39
Fixed Nodes in Narrative Puzzle Games: A Comparative Study of Storyteller and the WFP Relief Puzzle

ABSTRACT. This late-breaking paper compares the narrative structures of two puzzle-based storytelling games: Storyteller (Benmergui, 2023) and “Make Your Own Relief Puzzle,” a serious game commissioned by the World Food Programme (WFP) to communicate its 72-hour post-disaster relief principle. Both games ask players to assemble a story by manipulating discrete narrative units, yet they differ fundamentally in how fixed structural elements constrain player choice. We define two models to capture this difference: the Narrative Answer Model (NAM), in which a single fixed ending functions as the “correct answer” that all player combinations must converge on, and the Narrative Anchor Model (NAncM), in which a small number of Fixed Nodes function as narrative anchors that bound, rather than dictate, the player-authored story. Using Murray’s concept of procedural narrative and Bogost’s procedural rhetoric as an analytic lens, we show that the same puzzle-narrative combination can produce two opposite design philosophies — a game that finds the answer versus a game that builds the journey — depending solely on whether fixed elements function as an endpoint or as a boundary. We discuss the implications of NAncM for serious game design, where a client’s core message must survive open-ended player agency, report a small preliminary playtesting observation, and outline directions for validating the model across additional cases.