Towards a Learner-Centered Explainable AI: Lessons from the learning sciences
Project Overview
The document explores the integration of generative AI in education, emphasizing the design and evaluation of explainable AI (XAI) systems through a learner-centered approach that aligns AI with human learning objectives. It presents a framework aimed at systematically developing XAI to enhance understanding and decision-making among learners. A notable case study in AI-augmented social work illustrates the practical application of this framework, underscoring the importance of collaborative design involving various stakeholders to refine educational goals and improve the effectiveness of AI tools in authentic learning environments. The findings suggest that when AI systems are designed with a focus on human-centered learning, they can significantly enhance educational outcomes by fostering deeper comprehension and facilitating better decision-making processes. Overall, the document highlights the potential of generative AI to transform educational practices by ensuring that AI technologies are aligned with the needs and goals of learners, ultimately leading to more effective and engaging learning experiences.
Key Applications
Allegheny Family Screening Tool (AFST)
Context: AI-augmented social work, targeting social workers handling child maltreatment referrals.
Implementation: Collaborative design processes involving social workers to define learning objectives and create training materials.
Outcomes: Enhanced ability for social workers to understand and effectively integrate AI predictions into their decision-making processes.
Challenges: Initial lack of understanding of the AI tool among social workers, differing objectives between the AI predictions and workers' immediate decision-making needs.
Implementation Barriers
Organizational
Limited integration of AI tools into the daily workflows of social workers, leading to confusion and inefficiency.
Proposed Solutions: Co-design training materials with social workers to align AI outputs with their decision-making contexts.
Technical
Challenges in ensuring AI models align with the real-time needs and objectives of social workers.
Proposed Solutions: Iterative design and feedback processes that involve stakeholders in defining success metrics and learning objectives.
Project Team
Anna Kawakami
Researcher
Luke Guerdan
Researcher
Yang Cheng
Researcher
Anita Sun
Researcher
Alison Hu
Researcher
Kate Glazko
Researcher
Nikos Arechiga
Researcher
Matthew Lee
Researcher
Scott Carter
Researcher
Haiyi Zhu
Researcher
Kenneth Holstein
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Anna Kawakami, Luke Guerdan, Yang Cheng, Anita Sun, Alison Hu, Kate Glazko, Nikos Arechiga, Matthew Lee, Scott Carter, Haiyi Zhu, Kenneth Holstein
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai