Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Sketching AI Concepts with Capabilities and Examples: AI Innovation in the Intensive Care Unit

Nur Yildirim, Susanna Zlotnikov, Deniz Sayar, Jeremy M. Kahn, Leigh A. Bukowski, Sher Shah Amin, Kathryn A. Riman, Billie S. Davis, John S. Minturn, Andrew J. King, Dan Ricketts, Lu Tang, Venkatesh Sivaraman, Adam Perer, Sarah M. Preum, James McCann, John Zimmerman

TL;DR

This paper tackles the problem of ineffective early-stage ideation for AI in high-stakes healthcare settings. It introduces a three-phase, multidisciplinary design process using an ICU dataset, combining ideation, problem formulation with the Do-Reason-Know worksheet, and sketching/co-design with clinicians. Key contributions include a detailed case study of early-phase AI innovation in the ICU and a set of artifacts and methods—capability-based ideation, assessment matrices, and the Do-Reason-Know worksheet—that help identify low-risk, high-value AI concepts while surfacing data and ethical considerations. The findings show that starting from AI capabilities expands the design space, enables more feasible concepts, and improves stakeholder engagement, with implications for reducing AI project dropout in healthcare and other critical domains.

Abstract

Advances in artificial intelligence (AI) have enabled unprecedented capabilities, yet innovation teams struggle when envisioning AI concepts. Data science teams think of innovations users do not want, while domain experts think of innovations that cannot be built. A lack of effective ideation seems to be a breakdown point. How might multidisciplinary teams identify buildable and desirable use cases? This paper presents a first hand account of ideating AI concepts to improve critical care medicine. As a team of data scientists, clinicians, and HCI researchers, we conducted a series of design workshops to explore more effective approaches to AI concept ideation and problem formulation. We detail our process, the challenges we encountered, and practices and artifacts that proved effective. We discuss the research implications for improved collaboration and stakeholder engagement, and discuss the role HCI might play in reducing the high failure rate experienced in AI innovation.

Sketching AI Concepts with Capabilities and Examples: AI Innovation in the Intensive Care Unit

TL;DR

This paper tackles the problem of ineffective early-stage ideation for AI in high-stakes healthcare settings. It introduces a three-phase, multidisciplinary design process using an ICU dataset, combining ideation, problem formulation with the Do-Reason-Know worksheet, and sketching/co-design with clinicians. Key contributions include a detailed case study of early-phase AI innovation in the ICU and a set of artifacts and methods—capability-based ideation, assessment matrices, and the Do-Reason-Know worksheet—that help identify low-risk, high-value AI concepts while surfacing data and ethical considerations. The findings show that starting from AI capabilities expands the design space, enables more feasible concepts, and improves stakeholder engagement, with implications for reducing AI project dropout in healthcare and other critical domains.

Abstract

Advances in artificial intelligence (AI) have enabled unprecedented capabilities, yet innovation teams struggle when envisioning AI concepts. Data science teams think of innovations users do not want, while domain experts think of innovations that cannot be built. A lack of effective ideation seems to be a breakdown point. How might multidisciplinary teams identify buildable and desirable use cases? This paper presents a first hand account of ideating AI concepts to improve critical care medicine. As a team of data scientists, clinicians, and HCI researchers, we conducted a series of design workshops to explore more effective approaches to AI concept ideation and problem formulation. We detail our process, the challenges we encountered, and practices and artifacts that proved effective. We discuss the research implications for improved collaboration and stakeholder engagement, and discuss the role HCI might play in reducing the high failure rate experienced in AI innovation.
Paper Structure (43 sections, 5 figures, 2 tables)

This paper contains 43 sections, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Challenges of AI Product Innovation
  4. Broadening Participation in AI Design
  5. Designing AI for Healthcare
  6. Overview of Design Process
  7. The ICU Dataset
  8. Phase 1: Brainstorming AI Concepts
  9. Method
  10. Workshop 1: User-centered approach
  11. Workshop 2: User-centered and tech-centered approach
  12. Data Collection and Analysis
  13. Findings
  14. Workshop 1 Outcome
  15. Workshop 2 Outcome
  16. ...and 28 more sections

Figures (5)

  • Figure 1: An AI capability abstraction and example (left), poster printouts to prompt ideation across each capability (right).
  • Figure 2: Our first workshop resulted in ideas that were technically difficult, some of which were clinically relevant.
  • Figure 3: In the second workshop, concepts moved towards (a) low-effort and high-impact; (b) from high expertise-excellent performance to medium expertise-moderate performance.
  • Figure 4: Do-Reason-Know worksheet enabled us to detail each idea in terms of model reasoning, data, and interaction form.
  • Figure 5: Two low fidelity sketches detailing the concept of predicting if a mechanically ventilated patient is eligible to receive the SAT/SBT protocol. We used the sketches to conduct co-design workshops to elicit feedback from nurses and respiratory therapists.