Verbal engagement correlated more strongly with science learning in the interactive condition (response rate r = .485) than in the pseudo-interactive condition (r = .193)¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r3 · q3 · i3 large
Currency · newest study 2023 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i3 The correlation between response rate and science learning outcomes was significantly stronger in the interactive condition (r = .485, p < .001) than in the pseudo-interactive condition (r = .193, p = .09), with a statistically significant difference between correlations (p = .005). → Ying Xu 2023 q3 i3 Relevance correlated with learning at r = .585 in the interactive condition versus r = .269 in the pseudo-interactive condition, a statistically significant difference (p = .022); the accuracy correlation difference was not statistically significant (p = .09). → Ying Xu 2023 (2)

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 · i3 · causal · r3

Exploratory correlational analysis of four verbal-engagement metrics against post-test scores among the 156 children receiving questions (Figure 5, Table 5). Printed effect sizes: r = 0.485 (interactive) vs r = 0.193 (pseudo-interactive); "The difference in correlation between these two groups was statistically significant (p = .005)".

"For the interactive group, the correlation between response rate and science learning outcomes was 0.485 (p < .001), compared to a correlation of 0.193 (p = .09) for the pseudo-interactive group. The difference in correlation between these two groups was statistically significant (p = .005)"

Ying Xu 2023 (2)¶

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 · i3 · causal · r3

Same exploratory correlational analysis. Printed effect sizes: r = .585 (interactive) vs r = .269 (pseudo-interactive) for relevance–learning, "a statistically significant difference (p = .022)"; the accuracy correlation (r = .540 vs r = .377) did not differ significantly (p = .09).

"in the interactive condition, there was a higher correlation between relevance and science learning outcomes (r = .585, p < .001) compared to the pseudo-interactive condition (r = .269)—also a statistically significant difference (p = .022)"

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
• 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) — related
• Pseudo-interactive responses showed significantly more repetition: a lower type-token ratio than interactive responses (t = 3.371, p = .001) — 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