AVI 2020, Island of Ischia (Italy) & online - September 28, 2020
Even as computing makes transformative advancements in data-driven sub-disciplines such as AI and ML, and we witness unprecedented datafication of the society we live in, there is also growing awareness that emergent applications are systematically discriminating against many populations. A major driver of the bias is data, or the lack thereof. As long as the data are drawn using methods that align with a Western, universalist approach, gender, ethnic, racial, and cultural biases are likely to persist. Zou and Schiebinger have noted that over 45% of Imagenet data, which computer vision research draws from significantly, comes from the United States (US), where only 4% of the world's population resides. India and China together contribute only 3% of Imagenet data, while representing 36% of the world's population. The result is that a photograph of a traditional US bride is annotated accurately, while a photograph of a North Indian bride is recognized as 'performance art' and 'costume'.
Given this lack of representation in data, problems related to the robustness and representativeness of data infrastructures become more pressing. For data to be meaningful, they must be collected, stored, understood, analysed, and visualised, all from a holistic and contextually appropriate perspective. There may be challenges encountered in each of these stages; these challenges are exacerbated when we consider the cultural, technological, and/or infrastructural specificities of multilingually diverse and resource-constrained regions across the world. This is true for parts of the Global North as well as the Global South.
In many application domains such as global health, education, gender equality, agriculture, and others, the data burden is borne by workers from socioculturally and economically diverse backgrounds. Low digital expertise and different vantage points can mean that these workers lack the kind of data literacies required of them by their employers. Data-driven approaches can benefit from integration of a more human-centered orientation before being used to inform the design, deployment, and evaluation of technologies in various less-served contexts. These are some of the important conversations that our workshop seeks to advance.
We invite researchers and practitioners in interdisciplinary domains intersecting HCI, AI, ML, design, STS, and/or global development to engage in dialog on the topics above. A key priority of our workshop will be to invite submissions from an intellectually diverse and global group of participants to further discussions on how appropriate human-centered design can contribute to addressing data-related challenges among marginalised and under-represented/underserved groups around the world. In particular, we solicit participation across more and less technical researchers in HCI who are motivated to address the list of topics below.
Topics at the workshop must consider the integration of human-centered design in data-driven approaches, touching upon under-resourced contexts across the Global South and North. They include, but are not limited to:
Interfaces and Visualization:
Novel interfaces for deriving qualitative/quantitative insights from data
Interfaces to support data literacy in multilingual contexts
Information visualization tools and techniques for data literacy
Data literacy for end-users in under-resourced contexts
Data Infrastructures for Social Good:
Data collection and field research
Data quality
Data sharing
Privacy and transparency in data analytics
Human-centered design of data-driven approaches:
Interfaces for explainable AI
User-centered evaluations, techniques, and methods for AI and social good
Study of public concerns with AI-based technologies
Data work in specific application areas, such as:
Public/global health
Education
Agriculture
Gender equality
Refugee resettlement
and more.
We invite submissions of position papers in the CHI Extended Abstracts format, of 2-4 pages in length.
PDFs of submissions can be emailed to Luigi De Russis (one of the organizers) at luigi.derussis@polito.it. They will be reviewed by all organizers based on relevance, originality, and overall quality. Upon acceptance, at least one author of each accepted paper is required to attend the workshop. Virtual participation will be made possible.
All workshop participants must register for the workshop. Accepted and presented papers will be made available on CEUR Workshop Proceedings, if possible, while workshop results will be published on our website. Notifications will be mailed to the authors within 15 days of receipt (and no later than the date reported below).
Workshop results will be summarized and submitted as an article or blog post in Interactions or Communications of the ACM.
Workshop submissions due: July 31, 2020
Results announced: August 6, 2020
Camera-readies and registration due: August 21, 2020
Workshop date: September 28, 2020 (14:30-18:30 CEST)