*****************************************************************************************
IPMU 2022 –
Information Processing and Management of Uncertainty in Knowledge-Based Systems /
July 11-15, 2022 – Milan, Italy
https://ipmu2022.disco.unimib.it/
Special Session on “Data Perspectivism in Ground Truthing and Artificial Intelligence”
S3 – https://ipmu2022.disco.unimib.it/special-sessions/
**************************************************

 Description, scope and aims

Many Artificial Intelligence applications are based on supervised machine learning (ML), which ultimately grounds on manually annotated data. The annotation process (often called
ground-truthing) is often performed in terms of a majority vote and this has been proved to be often problematic, as highlighted by recent studies on the evaluation of ML models. Recently, a different paradigm for ground-truthing has started to emerge, called data perspectivism [1], which moves away from traditional majority aggregated datasets, towards the adoption of methods that integrate different opinions and perspectives within the knowledge representation, training, and evaluation steps of ML processes, by adopting a non-aggregation policy. This alternative paradigm obviously implies a radical change in how we develop and evaluate ML systems: such ML systems have to take into account multiple, uncertain, and potentially mutually conflicting views [2]. This obviously brings both opportunities and difficulties: novel models or training techniques may need to be designed, and the validation phase may become more complex. Nonetheless, initial works have shown that data perspectivism can lead to better performances [3,4], and could also have important implications in terms of human-in-the-loop and interpretable AI, as well as in regard to the ethical issues or concerns related to the use of AI systems [5]. Data perspectivism is a framework to treat uncertainty (the main theme of IPMU) at the level of knowledge modeling and its integration in the development and evaluation of systems.

The scope of this special session is to attract contributions related to the management of subjective, uncertain, multi-perspective, or otherwise non-aggregated data in ground-truthing, machine learning, and more generally artificial intelligence systems.
Invited contributions: full research papers and research in progress papers.

Topics of interest:

    Subjective, uncertain, or conflicting information in annotation and crowdsourcing processes;
    Limits and problems with standard data annotation and aggregation processes;
    Theoretical studies on the problem of learning from multi-rater and non-aggregated data;
    Participation mechanisms/incentives/gamification for rater engagement and crowdsourcing;
    Ethical and legal concerns related to annotation and aggregation processes in ground-truthing;
    Creation and documentation of multi-rater and non-aggregated datasets and benchmarks;
    Development of ML algorithms for multi-rater and non-aggregated data;
    Development of techniques to detect and manage multiple forms of uncertainty in multi-rater and non-aggregated data;
    Techniques for the evaluation of ML systems based on multi-rater and non-aggregated data;
    Applications of data perspectivism and non-aggregated data to interpretable, human-in-the-loop AI and algorithmic fairness;
    Experimental and application studies of ML/AI systems on multi-rater and non-aggregated data, in possibly different application domains (e.g. NLP, medicine,legal studies, etc.)

***************************************************
Important dates

Paper Submission deadline: Friday, 14 January 2022 Friday, 18 February 2022 (STRICT)

Valerio Basile (University of Turin, Italy),
Federico Cabitza (University of Milano-Bicocca, Italy)

***************************************************

Related readings

[1] Basile, V., Cabitza, F., Campagner, A., Fell, M. (2021)
Toward a Perspectivist Turn in Ground Truthing for Predictive Computing
arXiv preprint, arXiv:2109.04270
https://arxiv.org/pdf/2109.04270.pdf
[2] Zhang, J., Wu, X., Sheng, V.S. (2016)
Learning from crowdsourced labeled data: A survey.
Artificial Intelligence Review
[3] Fornaciari, T., Uma, A., Paun, S., Plank, B., Hovy, D., Poesio, M. (2021)
Beyond Black & White: Leveraging Annotator Disagreement via Soft-Label Multi-Task
Learning
Conference of the North American Chapter of the Association for Computational Linguistics:
Human Language Technologies (NAACL 2021)
[4] Campagner, A., Ciucci, D., Svensson, C.M., Figge, M. T., Cabitza, F. (2021)
Ground truthing from multi-rater labeling with three-way decision and possibility theory
Information Sciences
[5] Basile, V., Fell, M., Fornaciari, T., Hovy, D., Paun, S., Plank, B., Poesio, M., Uma, A. (2021)
We Need to Consider Disagreement in Evaluation
1st Workshop on Benchmarking: Past, Present and Future at ACL 2021