IBM Research AI focuses on Human-Centered Data Science at CSCW

Share this post:

Editor’s note: This post is written by Vera Liao, Michael Muller, Christine Wolf and Justin Weisz of IBM Research.

The ACM Conference on Computer-Supported Cooperative Work and Social Computing (CSCW) will be held November 9th-13th in Austin, TX. IBM Research AI is proud to be a Contributing sponsor of CSCW, the premier venue for research in the design and use of technologies that affect groups, organizations, communities, and networks.

Human-Centered AI at CSCW
IBM researchers will present three technical papers and one demo, participate in two panels, and organize four workshops at CSCW. These contributions reflect IBM’s recent research in advancing AI technologies that improve peoples’ workplace, organizational, and social experiences, including our participation in defining the emerging academic sub-discipline of Human-Centered Data Science (HCDS). Our work includes studies on:

  • How cutting-edge AI automation technologies and new visualizations impact the work practices of data scientists.
  • How machine learning technologies will transform the future of work in enterprise settings.
  • How to incorporate AI technologies in healthcare settings to improve clinicians’ work experience and patient outcomes, and;
  • How to develop an interdisciplinary research agenda for creating AI agents that better collaborate and communicate with humans.

Improving Data Science Work Practices
New technologies are being created that automate tedious or time-consuming tasks of data scientists. At IBM, our AutoAI capabilities make it easier for data scientists to automatically join disparate data sets and clean data, engineer features, select models, and tune hyperparameters. However, now that the tasks performed by human data scientists are being automated, what does it mean for the future of the field? IBM researchers investigated how data scientists perceive their job roles and the nature of their work in light of new AutoAI technologies. Despite concerns about the trend toward job automation, they found that data scientists held optimism toward a future in which AutoAI acts as a first-class collaborator in data science teams by helping automate labor-intensive tasks, guide analysis, and demonstrate best practices.

IBM researchers will demonstrate ModelLens, an interactive visualization system that helps data science teams monitor and improve their models. These tasks are becoming increasingly important as more business processes incorporate ML models. ModelLens supports data scientists by helping them identify, analyze, and annotate errors made by their models. The insights revealed by discovering how models make errors will enable data scientists to make improvements to their models to reduce error rate.

Collaborating With AI
AI technologies are becoming ubiquitous in work practices and everyday life. Not only will people increasingly need to make sense of AI actions, but they will also need to figure out how such systems can be collaborative partners in getting work done. IBM researchers will report on an empirical field study of a team developing an AI system in the IT services domain. Workers were supportive of the use of AI to transform their IT infrastructure design work, yet also raised a number of challenges in fully integrating AI outputs into their work practices. These findings add nuance and complexity to discourses of the “future of work,” highlighting the importance of considering the collaborative dimensions of human-AI interactions.

The panel “Human-Agent Communication: Connecting Research and Development in HCI and AI,” co-organized by an IBM researcher, will focus on AI agents with conversational interfaces (e.g., chatbot, virtual agents). The panel will bring together scholars from the AI and human-computer interaction communities to develop interdisciplinary research agenda advancing conversational AI that allows humans to communicate and work together with AI in a more natural fashion.

The workshop on Qualitative Methods for CSCW: Challenges and Opportunities addresses human analysis of data through multiple methods, including highly-rigorous approaches such as grounded theory. The workshop is an opportunity to explore how human qualitative analysis can complement data-driven quantitative analysis. Discussions at this workshop will contribute to our understanding of how to enhance machine learning with human-in-the-loop data-labeling, a core problem in Human-Centered Data Science.

Research on intelligent technologies that improve peoples’ organizational and social environment
Several papers, panels, and workshops address fundamental research questions, methods, and design concepts that can drive innovation in CSCW technologies.

IBM researchers investigated concepts and metaphors used in defining user requirements and designing services for smart and connected Internet of Things (IoT) environments. From an examination of domestic environments, we introduce “home worlds” as a design concept that draws attention to the large and diffuse social networks of people, artifacts, and information flows that shape a physical environment, and how such networks influence the design of IoT systems and services.

Advances in contemporary sociotechnical systems such as big data, AI and machine learning, and cloud computing call into question the ability of traditional design methods to adequately capture the actors, interactions, boundaries, scales, and scopes of processes or workflows that cross such dynamic, data-driven systems. Process representations such as workflow diagrams are important not only in the design of collaborative systems, but also the design of organizational workflows and services. The workshop “Mapping the “How” of Collaborative Action: Research Methods for Studying Contemporary Sociotechnical Processes” will bring together designers and researchers to address these questions and innovate design methods capable of mapping complex processes in contemporary sociotechical systems.

The workshop “Learning from Team and Group Diversity: Nurturing and Benefiting from our Heterogeneity” focuses on the notion that diversity in research teams is beneficial for the acceptance of papers at key conferences. This workshop will allow us to share initial findings from an internal study of diversity in authorship at IBM Research, as well as work with other scholars to define key challenges for diversity research in organizations.

human-centered data science

This post is written by (clockwise from left) Vera Liao, Michael Muller, Christine Wolf and Justin Weisz of IBM Research.

From human-human collaboration to human-AI collaboration
With decades of experience studying and improving human collaboration, IBM researchers are extending their experience in creating AI technologies that collaborate with humans. We aim to push the boundaries of AI technologies by augmenting them with human intelligence. We look forward to seeing you in Austin!

Human-AI Collaboration in Data Science: Exploring Data Scientists’ Perceptions of Automated AI
Dakuo Wang, Justin D. Weisz, Michael Muller, Parikshit Ram, Werner Geyer, Casey Dugan, Yla Tausczik, Horst Samulowitz, Alexander Gray

Evaluating the Promise of Human-Algorithm Collaborations in Everyday Work Practices
Christine T. Wolf, Jeanette Blomberg

Home Worlds: Situating Domestic Computing in Everyday Life Through a Study of DIY Home Repair
Christine T. Wolf, Kathryn E. Ringland, Gillian R Hayes

Human-Agent Communication: Connecting Research and Development in HCI and AI
Vera Liao, Yi-Chia Wang, Timothy Bickmore, Pascale Fung, Jonathan Grudin, Zhou Yu, Michelle Zhou

Collective Innovation
Elizabeth Gerber, Jeffrey V. Nickerson, Mira Dontcheva, Laura Dabbish, Charlie Hill

ModelLens: An Interactive System to Support the Model Improvement Practices of Data Science Teams
Yannis Katsis, Christine T. Wolf

Identifying Challenges and Opportunities in Human–AI Collaboration in Healthcare
Sun Young Park, Pei-Yi Kuo, Andrea Barbarin, Elizabeth Kaziunas, Astrid Chow, Karandeep Singh, Lauren Wilcox, Walter S Lasecki

Learning from Team and Group Diversity: Nurturing and Benefiting from our Heterogeneity
Michael Muller, Volker Wulf, Susan R Fussell, Ge Gao, Pamela J Hinds, Nigini Oliveira, Katharina Reinecke, Lionel Robert Jr, Kanya (Pao) Siangliulue, Chien-Wen Yuan

Mapping the “How” of Collaborative Action: Research Methods for Studying Contemporary Sociotechnical Processes
Christine T. Wolf, Julia Bullard, Stacy Wood, Amelia Acker, Drew Paine, Charlotte P Lee

Qualitative Methods for CSCW: Challenges and Opportunities
Andrea Forte, Shion Guha, Casey Fiesler, Jed Brubaker, Michael Muller, Nora McDonald

More AI stories

IBM researchers investigate ways to help reduce bias in healthcare AI

Our study "Comparison of methods to reduce bias from clinical prediction models of postpartum depression” examines healthcare data and machine learning models routinely used in both research and application to address bias in healthcare AI.

Continue reading

Pushing the boundaries of human-AI interaction at IUI 2021

At the 2021 virtual edition of the ACM International Conference on Intelligent User Interfaces (IUI), researchers at IBM will present five full papers, two workshop papers, and two demos.

Continue reading

Deep Document Understanding: IBM’s AI extracts data from complex documents

Using novel deep learning architectures, we have developed an AI that could help organizations, enterprises, and data scientists to easily extract data from vast collections of documents. Our technology allows users to quickly customize high-quality extraction models. It transforms the documents, making it possible to use the text they contain for other downstream processes such as building a knowledge graph out of the extracted content.

Continue reading