Search for: All records

Despite the importance of AI literacy for both children and adults, adults have been understudied. We developed short videos for adults that provided training on the basics of AI understanding, use, and evaluation. In an online experiment, 94 adults aged 30-49 were randomly assigned in a 1:2 ratio to view either short videos on AI history (control group) or AI literacy training videos (treatment group). The results showed that the intervention significantly improved people’s self-efficacy of AI use but not in AI understanding or evaluation. Interestingly, participants’ fears of AI bias, privacy violations, and job replacement increased after the training, although they remained below the midpoints. We argue that the heightened fear in the treatment group reflects a foundation for critical thinking skills, as it moves them closer to a more calibrated, moderate level of fear. Therefore, this study uniquely contributes by utilizing short-form experiential content to both educate and foster a more informed, critical interaction with AI technologies. The implications of designing AI literacy educational materials for adults were discussed. 

Though introduced nearly 50 years ago, the infinitesimal jackknife (IJ) remains a popular modern tool for quantifying predictive uncertainty in complex estimation settings. In particular, when supervised learning ensembles are constructed via bootstrap samples, recent work demonstrated that the IJ estimate of variance is particularly convenient and useful. However, despite the algebraic simplicity of its final form, its derivation is rather complex. As a result, studies clarifying the intuition behind the estimator or rigorously investigating its properties have been severely lacking. This work aims to take a step forward on both fronts. We demonstrate that surprisingly, the exact form of the IJ estimator can be obtained via a straightforward linear regression of the individual bootstrap estimates on their respective weights or via the classical jackknife. The latter realization allows us to formally investigate the bias of the IJ variance estimator and better characterize the settings in which its use is appropriate. Finally, we extend these results to the case of U-statistics where base models are constructed via subsampling rather than bootstrapping and provide a consistent estimate of the resulting variance.