Workshop at CHI 2026

Co-Data: Cultivating Effective Human–LLM Collaboration for Collaborative Data Processing

Co-Data brings together HCI, data management, and domain experts to explore how Large Language Models can collaborate with people in data-intensive workflows—beyond simple query–response interactions.

Where Barcelona, Spain · CHI ’26
When April 13–17, 2026 (half-day workshop)
View Call for Participation Contact organizers
Submission deadline: February 12, 2026 · 2–4 pages
Overview

Why Co-Data? Why now?

Modern data workflows are rarely solo efforts. They bring together people with different expertise, vocabularies, and degrees of trust in automation. At the same time, LLMs are increasingly embedded in tools for cleaning, integration, annotation, and querying of data.

Yet, their potential role as collaborators—helping align goals, translate between communities, and coordinate decisions in data work—remains largely unexplored.

This workshop invites the community to examine LLMs not just as tools, but as participants in collaborative, data-intensive workflows. We take Interdependence Theory as a starting point to reason about dependence, mutual responsiveness, and shared outcomes in human–LLM interaction, while deliberately comparing it with alternative lenses.

Topics & Themes

Focus areas

Human–LLM collaboration in data workflows

Data collection Cleaning & quality Integration Annotation Analysis & querying

We focus on data-intensive workflows where LLMs and humans jointly:

  • Acquire and curate data across heterogeneous sources.
  • Detect, explain, and resolve errors or inconsistencies.
  • Align entities and semantics across schemas and domains.
  • Produce and verify annotations and labels.
  • Formulate, refine, and interpret data queries and analyses in collaborative contexts.

Theoretical lenses & frameworks

Interdependence Theory Socio-technical Activity Theory Common ground Value-sensitive design

We explore how Interdependence Theory can be adapted, combined, or stress-tested against other frameworks to:

  • Characterize mutual dependence between humans and LLMs.
  • Model mediation roles vs. agentive roles for LLMs.
  • Inform both design and evaluation of collaborative systems.

Human & technical challenges

  • Working with structured data and domain-specific context.
  • Reliability, hallucinations, and bias in LLM outputs.
  • Data privacy and provenance in collaborative workflows.
  • Varying levels of data literacy and AI literacy among participants.
  • Cognitive load, trust calibration, and distribution of control.

Design & evaluation questions

  • How do we support mutual responsiveness and shared outcomes?
  • Which interaction patterns foster trust, fairness, and equity?
  • How do we measure coordination efficiency, shared understanding, and perceived agency?
  • When should LLMs act as mediators vs. active teammates vs. tools?
Program

Workshop structure (half-day)

The workshop combines short talks, interactive mapping exercises, and convergence sessions to move from examples to shared frameworks.

Duration Activity
15 min
Opening & framing
Welcome, goals, and a primer on Interdependence Theory for human–LLM interaction.
45 min
Author panel
Short talks from contributors showing how human–LLM collaboration appears in their domains.
30 min
Mapping session
Small-group activity to map Interdependence Theory principles to concrete collaboration scenarios.
Break
Coffee break & informal networking
40 min
Convergence I: Challenges & barriers
Joint identification of technical and human challenges (bias, privacy, cognitive load, agency, etc.).
40 min
Convergence II: Opportunities & frameworks
Sketching interaction patterns, evaluation rubrics, and decision trees grounded in interdependence.
10 min
Wrap-up & next steps
Collective summary, plans for shared resources, and future collaborations.
Participate

Call for Participation

We welcome 2–4 page submissions (position papers, work-in-progress, or case studies) that explore human–LLM collaboration in data-intensive, collaborative settings—from foundational theory and system design to empirical studies and domain applications.

Submission details

As data workflows grow increasingly complex and multidisciplinary, collaboration across diverse roles is often hindered by differences in expertise, vocabulary, trust, and goals. While Large Language Models (LLMs) are already used for data cleaning, integration, and querying, their potential to mediate and enhance collaborative data work remains underexplored. The workshop Co-Data: Cultivating Effective Human-LLM Collaboration for Collaborative Data Processing invites HCI researchers, data-management experts, and domain practitioners to examine LLMs as collaborators, not just tools, in team-based data workflows.
We seek contributions on the design and study of human-LLM collaboration in data-intensive settings. This includes novel interaction paradigms in which LLMs support collaborative workflows; mediate among stakeholders through aggregation, translation, or conflict resolution; and shape group dynamics around trust, control, and delegation. We are particularly interested in empirical studies and user research that investigate how LLMs influence coordination, shared understanding, and decision-making.
We also welcome system-level perspectives, such as multi-agent architectures for collaborative data work, as well as work addressing accessibility, explainability, and equity in human-LLM collaboration. In addition, we invite papers that mobilize complementary theoretical lenses (e.g., socio-technical, activity-theoretic, distributed cognition, coordination/common ground, value-sensitive design) to design and evaluate human-LLM interactions. Application-oriented studies are encouraged from domains such as healthcare, public policy, and scientific discovery, where cross-disciplinary collaboration is critical.

  • Interaction paradigms where LLMs support collaborative data workflows (collection, cleaning, integration, annotation, analysis).
  • LLMs as mediators between stakeholders (aggregation, translation, conflict resolution).
  • Effects of LLMs on coordination, shared understanding, trust, and delegation.
  • Multi-agent or system-level architectures for collaborative data work.
  • Accessibility, explainability, and equity in human–LLM collaboration.
  • Theoretical perspectives (e.g. socio-technical, activity theory, distributed cognition, common ground, coactive design, value-sensitive design).
  • Application-oriented studies (e.g. healthcare, public policy, scientific discovery).
Important Dates
Submission deadline 12th February 2026
Notification to authors 24th February 2026

Submissions via Google form. https://forms.gle/3kDuFKnozLmsgsQr7

Example topics

  • Interaction paradigms where LLMs support collaborative data workflows (collection, cleaning, integration, annotation, analysis).
  • LLMs as mediators between stakeholders (aggregation, translation, conflict resolution).
  • Effects of LLMs on coordination, shared understanding, trust, and delegation.
  • Multi-agent or system-level architectures for collaborative data work.
  • Accessibility, explainability, and equity in human–LLM collaboration.
  • Theoretical perspectives (e.g. socio-technical, activity theory, distributed cognition, common ground, coactive design, value-sensitive design).
  • Application-oriented studies (e.g. healthcare, public policy, scientific discovery).

At least one author of each accepted paper must attend the workshop. Accepted papers will be made available on the workshop website; we plan to publish proceedings in CEUR-WS and develop shared resources such as a Co-Data playbook, pattern library, and methods pack.

Team

Organizers

Amedeo Pachera
Amedeo Pachera
Université Claude Bernard Lyon 1, CNRS LIRIS (France)
Andrea Mauri
Andrea Mauri
Université Claude Bernard Lyon 1, CNRS LIRIS (France)
Kashif Imteyaz
Kashif Imteyaz
Northeastern University (USA)
Jie Yang
Jie Yang
TU Delft, Web Information Systems (Netherlands)
Eric Umuhoza
Eric Umuhoza
Carnegie Mellon University Africa (Rwanda)
Angela Bonifati
Angela Bonifati
Université Claude Bernard Lyon 1, CNRS LIRIS & IUF (France)
Michal Lahav
Michal Lahav
Google DeepMind (USA)
Nitesh Goyal
Nitesh Goyal
Google DeepMind (USA)
Accessibility & contact

Practical information

Contact

For questions about submissions, participation, or logistics, please contact the organizers: