Narrative, and in particular storytelling, is an important part of the human experience. Consequently, computational systems that can reason about narrative be more effective communicators, entertainers, educators, trainers. One central challenges reasoning generation, automated creation meaningful event sequences. There are many factors -- logical aesthetic contribute to success a artifact. Central this its understandability. We argue following two attributes narratives universal: (a)...

Interactive narrative is a form of digital interactive experience in which users create or influence dramatic storyline through their actions. The goal an system to immerse virtual world such that they believe are integral part unfolding story and actions can significantly alter the direction outcome story. In this article we review ways artificial intelligence be brought bear on creation systems. We lay out landscape about 20 years research explore successes as well open questions...

Humans are increasingly coming into contact with artificial intelligence and machine learning systems. Human-centered is a perspective on AI ML that algorithms must be designed awareness they part of larger system consisting humans. We lay forth an argument human-centered can broken down two aspects: (1) systems understand humans from sociocultural perspective, (2) help them. further argue issues social responsibility such as fairness, accountability, interpretability, transparency.

Story generation is the problem of automatically selecting a sequence events that meet set criteria and can be told as story. knowledge-intensive; traditional story generators rely on priori defined domain models about fictional worlds, including characters, places, actions performed. Manually authoring costly thus not scalable. We present novel class system generate stories in an unknown domain. Our (a) learns model by crowdsourcing corpus narrative examples (b) generates sampling from...

Automated rationale generation is an approach for real-time explanation whereby a computational model learns to translate autonomous agent's internal state and action data representations into natural language. Training on human can enable agents learn generate human-like explanations their behavior. In this paper, using the context of agent that plays Frogger, we describe (a) how collect corpus explanations, (b) train neural generator produce different styles rationales, (c) people perceive...

Narrative intelligence refers to the ability - human or computer organize experience into narrative. Recently, researchers have applied narrative create interactive systems, virtual worlds in which a story unfolds and user is considered character story, able interact with elements other characters world. The standard approach incorporating storytelling system script at design time. However, this limits system's adapt user's preferences abilities. alternative generate stories dynamically on...

This paper describes an approach for managing the interaction of human users with computer-controlled agents in interactive narrative-oriented virtual environment. In these kinds systems, freedom user to perform whatever action she desires must be balanced preservation storyline used control system's characters. We describe a technique, narrative mediation, that exploits plan-based model structure manage and respond users' actions inside world. define two general classes response situations...

Automated story generation is the problem of automatically selecting a sequence events, actions, or words that can be told as story. We seek to develop system generate stories by learning everything it needs know from textual corpora. To date, recurrent neural networks learn language models at character, word, sentence levels have had little success generating coherent stories. explore question event representations provide mid-level abstraction between and sentences in order retain semantic...

We introduce \em AI rationalization, an approach for generating explanations of autonomous system behavior as if a human had performed the behavior. describe rationalization technique that uses neural machine translation to translate internal state-action representations agent into natural language. evaluate our in Frogger game environment, training playing rationalize its action choices using A language corpus is collected from players thinking out loud they play game. motivate use...

Prithviraj Ammanabrolu, Mark Riedl. Proceedings of the 2019 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers). 2019.

Machine learning advances have afforded an increase in algorithms capable of creating art, music, stories, games, and more. However, it is not yet well-understood how machine might best collaborate with people to support creative expression. To investigate practicing designers perceive the role AI process, we developed a game level design tool for Super Mario Bros.-style games built-in designer. In this paper discuss our Morai Maker intelligent through two mixed-methods studies total over...

We present an unsupervised process to generate full video game levels from a model trained on gameplay video. The represents probabilistic relationships between shapes properties, and relates the stylistic variance within domain. utilize classic platformer Super Mario Bros. evaluate this due its highly-regarded level design. output in comparison other data-driven generation techniques via user study demonstrate ability produce novel more stylistically similar exemplar input.

Language-modeling--based approaches to story plot generation attempt construct a by sampling from language model (LM) predict the next character, word, or sentence add story. LM techniques lack ability receive guidance user achieve specific goal, resulting in stories that don't have clear sense of progression and coherence. We present reward-shaping technique analyzes corpus produces intermediate rewards are backpropagated into pre-trained order guide toward given goal. Automated evaluations...

The realm of Artificial Intelligence (AI)'s impact on our lives is far reaching – with AI systems proliferating high-stakes domains such as healthcare, finance, mobility, law, etc., these must be able to explain their decision diverse end-users comprehensibly. Yet the discourse Explainable (XAI) has been predominantly focused algorithm-centered approaches, suffering from gaps in meeting user needs and exacerbating issues algorithmic opacity. To address issues, researchers have called for...

Sarah Wiegreffe, Jack Hessel, Swabha Swayamdipta, Mark Riedl, Yejin Choi. Proceedings of the 2022 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies. 2022.

Many computer games of all genres pit the player against a succession increasingly difficult challenges such as combat with computer-controlled enemies and puzzles. Part fun is to master skills necessary complete game. Challenge tailoring problem matching difficulty skill-based events over course game specific player's abilities. We present tensor factorization approach predicting performance in games. Our data-driven can predict changes players' skill mastery time, allowing more accurate...

Explainability of AI systems is crucial to hold them accountable because they are increasingly becoming consequential in our lives by powering high-stakes decisions domains like healthcare and law. When it comes Explainable (XAI), understanding who interacts with the black-box just as important "opening" it, if not more. Yet discourse XAI has been predominantly centered around black-box, suffering from deficiencies meeting user needs exacerbating issues algorithmic opacity. To address these...

As AI-powered systems increasingly mediate consequential decision-making, their explainability is critical for end-users to take informed and accountable actions. Explanations in human-human interactions are socially-situated. AI often socio-organizationally embedded. However, Explainable (XAI) approaches have been predominantly algorithm-centered. We a developmental step towards socially-situated XAI by introducing exploring Social Transparency (ST), sociotechnically perspective that...

Explainable AI (XAI) systems are sociotechnical in nature; thus, they subject to the gap--divide between technical affordances and social needs. However, charting this gap is challenging. In context of XAI, we argue that improves our problem understanding, which can reflexively provide actionable insights improve explainability. Utilizing two case studies distinct domains, empirically derive a framework facilitates systematic by connecting guidelines XAI elucidating how use them address gap....

Explainability of AI systems is critical for users to take informed actions. Understanding who opens the black-box just as important opening it. We conduct a mixed-methods study how two different groups—people with and without background—perceive types explanations. Quantitatively, we share user perceptions along five dimensions. Qualitatively, describe background can influence interpretations, elucidating differences through lenses appropriation cognitive heuristics. find that (1) both...

The ability to generate narrative is of importance computer systems that wish use story effectively for entertainment, training, or education. We identify two properties - plot coherence and character believability which play a role in the success story. Plot perception by audience members actions have relevance outcome Character are motivated agents' internal beliefs desires. Unlike conventional planning plan goals represent an agent's intended world state, multiagent involves In order...