We invite research contributions to the Crowdsourcing and Human Computation track at the 32nd edition of The Web Conference series (formerly known as WWW), to be hosted at Austin, TX, US, on April 30 - May 4, 2023 ( https://www2023.thewebconf.org/)
Fifteen years ago, a 2007 WWW paper entitled “Internet-Scale Collection of Human-Reviewed Data” was one of several forerunners to signal a new, emerging area of research on Human Computation and Crowdsourcing (HCOMP). Growing excitement and work in this new area would eventually lead to four years of HCOMP workshops across KDD and AAAI (2009-2012), a new annual AAAI HCOMP conference (2013 onward), and a new, annual HCOMP track at the WebConference (2014 onward).
Today, the world and research landscape looks remarkably different than it did in 2007, with the Web playing a central role in orchestrating such advances. Of particular note, modern neural models have transformed AI capabilities, along with far greater ubiquity and significance of AI systems now in practical deployment around the world. As one effect of this, the commoditization and democratization of AI models today has also brought a new focus to “data-centric AI” in which AI models can succeed or fail based on the quality of underlying data and human annotations. The nature of human-AI interactions are also continually evolving in response to AI advances, posing an ever-changing frontier of new challenges for researchers and practitioners. Furthermore, the growth of AI power has brought a commensurate recognition of the need for responsible AI systems that are fair, accountable, transparent, and trustworthy – across diverse, global communities of human stakeholders who interact with or are impacted by AI systems. Given the central role of HCOMP in AI (creating reliable training and benchmark annotations, as well as enabling hybrid, human-in-the-loop systems), continuing innovation in HCOMP remains a key challenge for the further advancement of AI. HCOMP itself has made tremendous strides forward in the past fifteen years, yet many research challenges remain.
Our track invites AI, HCI, and related contributions that advances the broad spectrum of crowdsourcing and human computation (HCOMP) in the scope of the Web:
- algorithms, analysis, applications, methods, systems, and techniques
- conceptual, empirical, theoretical, and mixed-methods
- spanning fields (e.g, psychology, sociology, economics, ethics, etc.)
- system-centered, human-centered, and hybrid
More specifically, we invite work addressing contemporary HCOMP challenges including (but not limited to) the following Web-related themes:
Fundamental research challenges in Web-based HCOMP
- Data collection, generation, labeling, and cleaning: data-centric AI; human and AI-assisted annotation; annotator agreement, aggregation, and modeling; annotation subjectivity and ambiguity, data excellence; human-in-the-loop data augmentation, generation, and adversarial attacks; label noise and bias detection and reduction; task decomposition, task and workflow design, novel modalities for input acquisition, etc.
- Human-centered explainability : algorithmic/model explanations, interpretability, and transparency to enhance human success in using AI in decision-making, model and data debugging, task performance, trust in AI systems, appropriate reliance, etc. (please also read the CFP of the “Fairness, Accountability, Transparency and Ethics” track)
- Human-centered studies: collaborative systems, computer-supported cooperative work, human-computer interaction, human factors, interaction design, usability, user experience, etc.
- Resources, benchmarking, reproducibility: New resources for the community (e.g., datasets, open source toolkits, etc.), benchmarking studies comparing state of the art methods, and/or reproducibility studies of prior work.
- Addressing bias and diversity in annotation and human computation: Methods and algorithms to identify and mitigate biases in annotations; bias-aware annotation workflows; diversity in annotators and workers, data labeling, and hybrid, human-in-the-loop systems; downstream effects of annotator diversity on bias and fairness measures; impact on evaluation of various systems (e.g. information retrieval systems, recommender systems, etc.); ethics and fairness of HCOMP practices
Underlying workforce powering Web-based HCOMP
- Social and economic impacts of human computation and crowdsourcing: societal and methodological challenges around crowdsourcing labor and workforces; inequalities in access and representation in crowdsourcing workforces; platform affordances and economic impact
- Supporting HCOMP workers: collective action; design activism; fair work; ghost work, heteromation, and invisible work; human computation, digital colonialism, and the global south; impact sourcing and responsible sourcing; regulation; worker empowerment, organization, protection and wellness; and workforce diversity, equity, and inclusion, etc.
- Future of work: AI-assisted human coordination, team formation and work, distributed work, freelancer economy, hybrid, human+AI work and complementarity, etc.
Web-based HCOMP systems, frameworks, or architectures
- Crowd-powered systems: data management, marketplace design and sustainability, platforms, scalability, security, privacy, programming languages, real-time crowdsourcing, etc.
- Human-in-the-loop architectures: decision support; human-AI collaboration, interaction, and teaming; hybrid systems; mixed-initiative design, etc.
- Crowdsourcing: citizen science, collective intelligence, crowd computing, crowd creativity, crowdfunding, crowd ideation, crowd intelligence, crowd sensing, crowdsourcing contests, crowd phenomena, crowd science, incentive schemes, gamification, human flesh search, open innovation, peer production, prediction markets, reputation systems, social web, wisdom of crowds, etc.
- Human computation: decision-theoretic and game-theoretic design, design patterns, human algorithm design and complexity, mechanism and incentive design, etc.
Web-based Applications of HCOMP
- Machine learning for HCOMP: aggregation, answer fusion, annotator and user modeling, quality assurance, optimization, task assignment and recommendation, truth inference, etc.
- New Applications and Services: delivering beyond state-of-the-art AI capabilities and enhanced services through human computation and human-in-the-loop systems.
Authors should consult the conference’s main Research Track CFP to ensure their submissions are aligned with broader conference expectations, scope, and theme: “Web Research with Openness, Fairness and Reproducibility”. The CFP also details submission guidelines, relevant dates, and important policies. Review criteria will include considerations typical of those in past years of this track and the AAAI HCOMP conference, e.g., https://www.humancomputation.com/2016/review-criteria.html.
Submissions that are out of scope or unresponsive to the call above will be rejected early during the reviewing process (“desk rejected”) with minimal feedback.This includes submissions that:
- merely apply HCOMP methods in standard, previously known ways, without novel contributions to advance the methodology itself;
- do not relate to the web or web-based human computation platforms, methods, or applications.
In case you have doubts whether your paper fits the scope of this track, please contact the track chairs firstname.lastname@example.org