Outcomes
We in contrast Wise Agent to a standard, voice-controlled AR assistant baseline. We measured cognitive load utilizing the NASA Process Load Index (NASA-TLX), total usability with the System Usability Scale (SUS), consumer choice on a 7-point Likert scale, and complete interplay time.
Probably the most vital discovering was the discount in cognitive workload. The NASA-TLX knowledge confirmed that on a 100-point scale for psychological demand, the typical rating for Wise Agent was 21.1, in comparison with 65.0 for the baseline with a statistically vital distinction (𝑝 < .001). We noticed an analogous vital discount in perceived effort (𝑝 = .0039), which means that the proactive system efficiently offloaded the psychological work of forming a question.
Concerning usability, each techniques carried out properly, with no statistically vital distinction between their SUS scores (𝑝 = .11). Nonetheless, individuals expressed a robust and statistically vital choice for Wise Agent (𝑝 = .0074). On a 7-point scale, the typical choice score was 6.0 for Wise Agent, in comparison with 3.8 for the baseline.
For the interplay time, logged from the second a immediate was triggered to the ultimate system response to the consumer’s enter, the baseline was sooner (μ = 16.4s) in comparison with Wise Agent (μ = 28.5s). This distinction is an anticipated trade-off of the system’s two-step interplay circulate, the place the agent first proposes an motion and the consumer then confirms it. The sturdy consumer choice for Wise Agent suggests this trade-off was acceptable, notably in social contexts the place discretion and minimal consumer effort had been essential.
