AIfred and ChatGPT produce comparable math scores while assistance is available¶

Claim · All claims
Evidence · 1 study · causal r1 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? With assistance available, math assignment scores did not differ significantly between AIfred (6.7/10) and ChatGPT (7.3/10); equivalence was not tested. → Gregorio Orlando 2026

Evidence¶

Gregorio Orlando 2026¶

Gregorio Orlando, Milan Groshev, Eduardo Castelló Ferrer. (2026). AIfred: Augmented Learning through Functional Robotic Embodiment at the Desk. https://arxiv.org/abs/2609.38737 doi:10.48550/arxiv.2609.38737

q2 · i? · causal · r1

In the assisted math assignment, graded by four anonymized Gemini grading agents on an eight-criterion rubric (mean inter-agent SD 0.53), scores "do not differ significantly between AIfred and ChatGPT" (6.7/10 vs. 7.3/10; p=.41). No effect size is printed.

"participants' task scores do not differ significantly between AIfred and ChatGPT (6.7/10 vs. 7.3/10;p=.41)."

Discussion¶

Related Claims¶

• Robotic projected AI assistance yields 60% higher short-term learning transfer than screen-based ChatGPT once assistance is withdrawn — related
• AIfred reduces observed physical-digital context switches by 98% relative to screen-based assistance, while task completion time is longer — related
• AIfred users report higher perceived learning support, innovation, satisfaction, and cognitive demand, with no difference in perceived productivity — related
• Student effort on assignments, measured by time-on-task and keystrokes, showed no evidence of decrease when AI use was allowed — related
• Learners receiving ChatGPT-generated hints showed statistically significant learning gains compared to a no-hint control in an online mathematics tutoring experiment — related