The Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (IAAI-21) is a venue for papers describing highly innovative realizations of AI technology. The objective of the conference is to showcase successful applications and novel uses of AI. The conference will use technical papers, best practice papers, invited talks, and panel discussions to explore issues, methods, and lessons learned in the development and deployment of AI applications; and to promote an interchange of ideas between basic and applied AI and the discourse on the actual deployment of AI in practice.

IAAI-21 will consider: (1) papers that showcase novel, deployed applications of AI, and potential applications on this trajectory; (2) papers that present tools for faster AI solutions development and deployment; (3) papers that showcase original ways of integrating methodologies from different areas of AI for practical realization; as well as (4) best practice papers. Submissions should clearly identify which track they are intended for, as the tracks are judged on different criteria. All submissions must be original.

Tracks and Topics

1. Highly Innovative Applications of AI

Papers submitted to this track must describe deployed applications with measurable benefits that include an innovative use of AI technology. Applications are defined as deployed once they are in production use by their final end-users and the in-use experience can be meaningfully collected and reported. The study may evaluate either a stand-alone application or a component of a complex system.

Papers will be judged primarily by the quality of: the task or problem description; the application description; the innovative use of AI technology; the application use and payoff; and the lessons learned during application development, deployment and maintenance.

Original papers on the aspects of deploying AI applications in practice are welcome, and papers, while expected to exhibit both innovative use of AI as well as demonstrated impact, may focus more on one of these aspects. Each accepted deployed application paper will receive the IAAI ‘Innovative Application’ Award.

Papers in this track may have up to 8 pages in the prescribed AAAI style, plus at most one more page which may only contain references.

2. Emerging Applications of AI

Emerging applications papers ‘bridge the gap’ between basic AI research and case studies of deployed AI applications, by discussing efforts to apply AI tools, techniques, or methods to real-world problems in novel ways. Emerging applications focus on aspects of AI applications that are not yet sufficiently deployed to be submitted as case studies in the first track.

This track is distinguished from reports of purely scientific AI research appropriate for the AAAI-21 Conference in that the objective of the efforts reported here should be the potential application of AI technologies, including engineering considerations. A requirement for papers is to discuss the path forward for achieving deployment of the technology.

Papers will be judged primarily by the following criteria: significance (of the problem, and the tool or methodology); relevance of AI technology to the problem; innovation; path to deployment; content; evaluation; technical quality; and clarity. Authors are advised to bear these questions in mind while writing their papers.

Papers in this track may have up to 6 pages in the prescribed AAAI style, plus at most one more page which may only contain references.

3. Innovative Tools for Enabling AI Application

Within this track, we solicit papers describing tools to improve applied AI innovation and deployment of AI systems. Areas of interest include, but are not limited to:

    Process organization: Tools that help manage and assure the development, evaluation or deployment of AI systems.
    Data cleaning: Tools to ease the pain point of processing raw data for its use in AI systems.
    Modelling: Tools that facilitate AI modelling, for example approaches to facilitate deriving models from examples or demonstrations.
    Meta-algorithmics: Tools to improve AI systems, algorithm configurators, algorithm portfolios, and hyper-parameter tuners.
    Platforms: Tools in the form of platforms designed for AI research, especially those with connections to real world systems and domains.
    Novel computational models: Tools to exploit new computational hardware, for example neuromorphic co-processors, quantum computers, and chips approximating quantum computation.
    Interaction: Tools that facilitate the interaction design of or user experience with AI systems, or improve user evaluations.
    Trusted AI: Tools that generate explanations, justifications, and persuasions for data-driven models.

Papers will be judged primarily by the following criteria: the extent to which the tool presented yields solutions with demonstrated improved quality; lower development, deployment and maintenance costs; better productivity; fewer errors; and higher ability to scale. Reviewers will pay attention to how well designed is the tool, how easy is it to use and how well is it documented, and how many users it has. Note the focus is on the tool, not on foundational research.

Papers in this track may have up to 6 pages in the prescribed AAAI style, plus at most one more page which may only contain references.

4. Innovative Inter-disciplinary AI Integration

This track is devoted to the integration of AI components with the focus on how the orchestration of methods from different AI silos requires the adaptation of existing technologies to allow them to work together well for application of AI in practice. Papers must pay attention to engineering considerations and, where relevant, to human–computer interaction.

Examples for such orchestrated new capabilities and applications include but are not limited to:

    Learning search algorithms: Implementations of systematic and heuristic search algorithms that are capable to improve their performance by learning from experience or during search.
    Decision support under uncertainty: Implementations that combine statistical and deterministic reasoning to provide scalable decision support under uncertainty for applications that are inherently stochastic.
    Knowledge representation: Representations that work effectively for downstream analytics, e.g., the innovative use of a knowledge graph for feature generation of a downstream machine learning model.
    Auto-configuration: The adaptation of search methods from combinatorial optimization to optimize knowledge bases or to design superior forecasting algorithms.
    Other research prototypes that integrate algorithms and methods from traditionally different AI sub-communities.

A clear, transparent, and reproducible case must be presented that allows the community to judge objectively the value of the innovation of the integration when compared to existing or ad-hoc approaches. Computational experiments on benchmarks (that either already are or will be made public) and in comparison with existing state-of-the-art baseline methods are expected. At least a portion of the experiments is expected to consider real-world (not synthetic generated) benchmarks to demonstrate the practical importance of the problem studied.

Papers in this track may have up to 6 pages in the prescribed AAAI style, plus at most one more page which may only contain references.

5. AI Best Practices, Challenge Problems, Training AI Users

In this final short paper track, we welcome papers that review best practices when deploying AI, that communicate novel challenge applications, and that review contributions that lower the barrier to applying AI by practitioners outside the AI community. These papers will be reviewed based on different criteria than the longer papers of the main IAAI tracks. Best practice papers must tie the recommendations to concrete learning from prior experience when deploying AI methods. Challenge problem papers must (this is a hard requirement) make non-generated, real-world benchmarks publicly available. AI training papers must focus on the particular challenges when bridging knowledge from application domain experts and AI expertise and how the domain experts can effectively and efficiently learn enough about the AI tools they use to apply them successfully. For all topics in the scope of this track, papers that challenge the status quo are particularly welcome.

Papers in this track may have up to 4 pages in the prescribed AAAI style, including references.