NYU Researchers Discover Longer AI Processing Delays Increase Perceived Quality Among 240 Study Participants

NYU study finds that users trust AI more when it takes longer to respond, interpreting technical latency as careful deliberation and deeper thought.

By: AXL Media

Published: Apr 25, 2026, 9:43 AM EDT

Source: Information for this report was sourced from EurekAlert!

The Illusion of Machine Deliberation

Recent findings presented at the CHI’26 conference indicate that the industry-wide obsession with eliminating AI latency may be overlooking a fundamental aspect of human psychology. While developers traditionally view the gap between a prompt and a response as a technical hurdle to be minimized, researchers Felicia Fang-Yi Tan and Oded Nov discovered that these pauses serve as social cues. According to the study, users do not view AI as a purely deterministic machine but rather as a conversational partner, often interpreting a few seconds of silence as the system engaging in deep thought or careful consideration.

Empirical Testing of Synthetic Latency

To investigate the impact of response timing, the research team recruited 240 individuals to perform a variety of knowledge-based tasks ranging from creative brainstorming to analytical advice-seeking. The experimental environment was strictly controlled to deliver responses at fixed intervals of two, nine, or twenty seconds, regardless of the complexity of the query. By isolating the variable of speed, the study aimed to determine if the physical wait time altered how the underlying intelligence was judged by the human operator, even when the quality of the text remained identical across all test groups.

Perception Diverges From Behavioral Patterns

The data suggests a fascinating disconnect between how people act and how they feel during an AI interaction. Surprisingly, the speed of the interface did not significantly alter user behavior, as participants prompted and refined their queries at similar frequencies regardless of the delay. Instead, the nature of the task itself dictated the interaction style, with creative drafting requiring more back-and-forth iteration than advice-focused queries. According to Tan, the findings prove that while timing does not change how people use the tool, it fundamentally shifts how they perceive the value of the information provided.