Presentation
Enhancing Statistical Numeracy Measurement Through a Human-in-the-Loop, AI-Assisted Item Development Approach
SessionWednesday Poster Session
DescriptionRisk misunderstanding is a costly and preventable cause of decision-making errors for experts and non-experts alike (Cokely et al., in press; Gigerenzer et al., 2007; Reyna et al., 2009). While numerous factors contribute to decision vulnerability, tests of statistical numeracy like the Berlin Numeracy Test have been found to be among the strongest general predictors of decision quality and risk literacy (e.g., the ability to evaluate and understand risk; Cokely et al., 2012; 2018). Recently, some researchers have suggested there is a need to expand the pool of validated numeracy questions across the range of difficulty (≈ 90 items) to support advanced applications (e.g., personalized training and assessment protocols; enhance precision of IRT-based latent-skill measurement models). Can traditional and structured human-in-the-loop, AI-assisted workflows be used to generate valid and robust items to expand the Berlin Numeracy Test item pool?
Event Type
Poster
TimeWednesday, October 21st5:30pm - 6:30pm PDT
Location
Similar Presentations