UIBK 2019: Workshop on User Interactions for Building Knowledge

Country: USA

City: Los Angeles

Abstr. due: 14.12.2018

Dates: 17.03.19 — 20.03.19

Area Of Sciences: Technical sciences;

Organizing comittee e-mail: jtchrist@ncsu.edu

Organizers: Marriott Marina Del Rey (at ACM IUI 2019)


Building sets of complete, correct, and unbiased information – whether it is domain knowledge or training data – is an iterative and ongoing process that is necessary for producing systems that have the requisite knowledge to be effective in their environment. Currently, this is an unintuitive task for users who have little to no knowledge of how the system and its underlying algorithms function. But splitting the process between AI-experts who understand the system and subject matter experts who understand the domain is inefficient. Under this model, production systems that need to keep up with rapid changes in the real world where they are meant to operate cannot be maintained efficiently and stagnate. If AI is ultimately to be effective, new platforms and interaction methods for completing these tasks that can be managed entirely by AI-novice stakeholders will be necessary.

We are soliciting submissions that address any area of guiding novice users through training or model building, including but not limited to:

  •     curating the set of knowledge or training data (e.g., helping users understand what the system does and does not know)
  •     obtaining knowledge from users of various levels of expertise
  •     managing/mitigating human bias in the final system
  •     the role of explainability (XAI) during training/model building
  •     the effect of human generated knowledge in the explainability of the system's decisions
  •     interactions for exchanging complex knowledge (fuzziness, multi-classification, etc.)
  •     refining systems from a pre-built model (maintenance, personalization)
  •     considerations around soliciting/collecting information from end users (ethics, error checking, data poisoning, etc)
  •     interaction modalities and novel visualizations for communicating model state to end users

We particularly encourage contributors to address and illustrate issues like these with case studies that explore the training/model building issues in specific subfields of AI such as training for machine learning classification, domain construction for automated planning, model refinement for data mining and more. This workshop seeks to bring together researchers and practitioners across different AI spaces to share and discuss challenges of and proposed solutions for building interactions for guiding novice users through knowledge collection and model building or training.

Submission Guidelines

Conference Web-Site: https://easychair.org/cfp/UIBK2019