Novelty Effect

• Category: Control › Human in the loop
• Tags: human-robot-interaction, user-studies, long-term-interaction, evaluation, social-robots

The change in people's behavior and ratings caused by a robot being new to them, which typically fades or shifts as the robot becomes familiar over repeated interactions.

Definition

The novelty effect is the influence that the newness of a technology has on how people respond to it. When people first meet a robot, they are often more curious, more engaged and more positive than they will be later, simply because it is unfamiliar. As the robot becomes familiar, interaction rates, attention and ratings commonly drop, and what remains reflects the robot's actual usefulness and appeal. Novelty can also work in the other direction, producing early anxiety or caution that eases with exposure.

In human-robot interaction, the novelty effect is a central threat to the validity of short studies. Many experiments involve a single session of minutes, so the measured engagement, trust or preference may reflect the first encounter rather than how the robot will be used day to day. Long-term field studies with social robots in schools, homes and workplaces have repeatedly found high early use that declines over the following days or weeks. Behaviors that seem charming at first, such as jokes, greetings or small talk, can become repetitive, while practical benefits may only show after people have learned how to use the robot.

Researchers manage the novelty effect in several ways: running multiple sessions or longer deployments, adding familiarization sessions before measurement, analyzing trends over time rather than single snapshots, varying robot behavior to reduce repetition, and comparing against participants who already have experience with robots. Reporting how long and how often participants interacted with the robot helps readers judge how much novelty may have shaped the results.

Why it matters

Engineers and designers often make decisions from short user studies. If those results are inflated or distorted by novelty, a robot or feature that tested well may disappoint once deployed. Recognizing the novelty effect leads to better study design and more realistic expectations for long-term use.

Key applications

• Designing and interpreting HRI user studies
• Long-term deployments of social and service robots
• Educational robots in schools
• Evaluating engagement features such as small talk and personality
• Product testing of consumer and home robots

In practice

A team adds small talk to a robot that works alongside people on an assembly task. In the first session participants rate the chatty robot much higher than a silent version. Instead of stopping there, the team runs five sessions over two weeks with the same participants. Ratings for the chatty robot decline after the second session, and logs show participants responding less to its remarks, while task performance does not change. The team keeps a reduced set of varied small-talk behaviors, introduces new topics over time, and reports the full trend rather than only the first-session result.

Common misconceptions

• The novelty effect always makes results more positive — it can also cause initial anxiety or caution that fades over time
• A single long session removes novelty — familiarity usually builds over repeated encounters across days, not only within one session
• Declining engagement means the robot failed — some decline is expected as novelty wears off, and the level that remains is the more useful measure

Who should know this

• HRI researchers planning studies and deployments
• Designers of social, educational and home robots
• Product teams interpreting user testing results

Related technologies

• Longitudinal Studies
• Within-Subjects Study Design
• Field Deployments
• Interaction Logging

Related terms

• Godspeed Questionnaire
• Perceived Safety
• Research Through Design
• Humanoid Robot
• End-User Robot Programming

Related projects

• evo: De-facto standard tool for evaluating and plotting odometry/SLAM trajectories
• Reachy Mini: Open-source desktop robot from Pollen and Hugging Face with an expressive 6-DoF head
• RoboDojo: Manipulation benchmark that runs in Isaac Sim and on real arms
• Inspect Robots: Define a robotics benchmark once, run any policy on any robot
• SenseNova-RoboRSI: Open runtime that runs, scores and improves robot agents driven by language models
• RAI: Vendor-agnostic agent framework for ROS 2 robots and physical AI