Contingent AI interaction increased children's likelihood of responding over time (OR = 1.026 per question), while pseudo-interactive interaction decreased it (OR = 0.96)¶

Claim · All claims
Evidence · 1 study · causal r3 · q3 · i0 negligible
Currency · newest study 2023 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i0 In a multi-level logistic regression, each additional question in the interactive condition increased the odds of a child verbally responding by 2.6% (OR = 1.026, p = .01), while in the pseudo-interactive condition the odds decreased by 1.4% (OR = 0.96, p = .005). → Ying Xu 2023

Evidence¶

Ying Xu 2023¶

Ying Xu, Kunlei He, Julian Levine, Daniel Ritchie, Zexuan Pan, Andres Bustamante, and Mark Warschauer. (2023). Artificial Intelligence Enhances Children's Science Learning from Television Shows. https://osf.io/j7upd

q3 · i0 · causal · r3

Multi-level logistic regression on the 156 children in interactive (n = 79) and pseudo-interactive (n = 77) conditions; printed odds ratios are OR = 1.026 and OR = 0.96, both negligible in magnitude. The article concludes "contingent interaction led to children steadily increasing their verbal engagement, whereas pseudo-interaction led to a steady decrease in engagement" (Figure 4).

"A multi-level logistic regression analysis revealed that in the interactive condition, for each additional question, the odds of a child verbally responding increased by 2.6% (OR = 1.026, p = .01). In contrast, in the pseudo-interactive condition, for each additional question the odds of a child responding decreased by 1.4% (OR = 0.96, p = .005)"

Discussion¶

Related Claims¶

• Contingent AI feedback outperformed pseudo-interactive generic feedback: interactive viewers scored 0.16 standard deviations higher than pseudo-interactive viewers on science learning — related
• Verbal engagement correlated more strongly with science learning in the interactive condition (response rate r = .485) than in the pseudo-interactive condition (r = .193) — related
• Children in the interactive condition responded to 77.94% of Elinor's questions versus 68.13% in the pseudo-interactive condition — related
• Pseudo-interactive videos also outperformed non-interactive videos by 0.16 standard deviations on science learning — related
• Pseudo-interactive responses showed significantly more repetition: a lower type-token ratio than interactive responses (t = 3.371, p = .001) — related