Robot Demonstration Videos Should Include Transparency Disclosures

Nozomi NakajimaarXiv 2026 · arXiv

TL;DR

Robot demonstration videos are the primary way the public perceives progress in AI, yet most viewers cannot distinguish between autonomous feats, human teleoperation, or even CGI. This paper demonstrates through a massive six-country study (N=9,341) that the lack of transparency leads to a dangerous overestimation of robot capabilities, and proposes that simple, standardized text disclosures are a necessary "gold standard" to calibrate public expectations and ensure safety.

What Problem Does This Paper Address?

In the "hype cycle" of modern robotics, a viral video can move markets and shape policy. However, unlike Large Language Models (LLMs) which anyone can test via a browser, "embodied AI" (robots) remains physically and financially inaccessible to the general public. Consequently, our collective understanding of what robots can actually do is filtered through highly curated demonstration videos.

The authors identify a critical "transparency gap": many high-profile videos omit whether a robot is being teleoperated by a human or if the footage has been digitally altered. This isn't just a matter of academic honesty; it creates a distorted mental model in the public consciousness. When people overestimate a robot's autonomy, the stakes range from investor disillusionment to actual physical harm (e.g., a safety driver over-relying on a vehicle’s automated systems).

Methods

The researchers employed a two-stage empirical approach:

  1. Expert Wiki Survey: They first surveyed 107 robotics experts (with 10+ years of experience) to identify which features are most important for public transparency. Using a "catch-up" algorithm to build consensus, they identified two primary dimensions: Visual Authenticity (Real vs. CGI) and Behavioral Authenticity (Autonomous vs. Teleoperated).
  2. Large-Scale Randomized Experiment: They conducted a pre-registered study with 9,341 participants across six countries (Brazil, China, France, India, South Korea, and the US). Participants were shown 12 robot videos from a bank of 20 stimuli.
  3. Control Group: Saw videos with no extra info.
  4. Treatment Group: Saw videos with a simple text disclosure (e.g., "The movement of the arm is actively remote controlled by a human operator").
  5. CGI Generation: To test visual discernment, the authors used stable diffusion-based image-to-video models to create physically implausible but visually "clean" robot videos.

Datasets

  • Expert Sample: 107 roboticists recruited from the 2024 Robotics Summit & Expo.
  • Public Sample: 9,341 adults, nationally representative across six countries.
  • Video Stimuli: A curated bank of 20 videos featuring various form factors (quadrupeds, arms, humanoids) performing manipulation and mobility tasks. This included "ground truth" labels for teleoperation and CGI.

Key Insights and Ideas

  • The "Coin Flip" of Autonomy: Without disclosures, the public’s ability to identify teleoperation is only 52.4%—essentially no better than a random guess. People are slightly better at identifying real footage (74.8%) but abysmal at spotting CGI (36.5%).
  • Overestimation is the Default: The lack of disclosure doesn't just cause confusion; it specifically causes overestimation. Viewers assume robots are more capable of navigating real-world environments than they actually are.
  • Disclosures Work (Mostly): Adding a simple text tag significantly improved the public's ability to discern truth. However, while the "vigilance" (discernment) persisted slightly after the disclosures were removed, the "calibration" (accurate estimation of what the robot can do) vanished. This suggests that disclosures must be persistent to be effective.
  • Robotics is Unique: Unlike "deepfakes" of humans where we have strong biological baselines for what looks "right," the public has no daily baseline for how a robot should move, making them uniquely vulnerable to deceptive editing.

Key Results

  • Visual Discernment: Disclosures improved the accuracy of identifying CGI vs. real footage by 4.7 percentage points (a 7.6% relative improvement).
  • Behavioral Discernment: Disclosures improved the identification of teleoperation by 2.9 percentage points (a 5.5% relative improvement).
  • Capability Accuracy: Disclosures increased the correct identification of environmental capability by 1.5 percentage points and task capability by 1.2 percentage points.
  • The Driver of Change: Regression analysis showed that these improvements were driven almost entirely by correcting overestimation rather than fixing underestimation.

Limitations and Future Work

  • Stimulus Constraints: The study relied on a specific set of videos from one institute. Future work should test if these findings hold across a broader range of corporate and academic "hype" videos.
  • The "Social Media" Factor: A survey environment is controlled. The authors acknowledge that the "demand effect" (participants knowing they are in a study) might differ from how people react while scrolling TikTok or X (formerly Twitter).
  • Standardization: While the paper proposes a "minimal viable disclosure" (a persistent text tag in the corner), the actual industry-wide implementation would require a multi-stakeholder effort involving journals (like IEEE), conferences, and social media platforms.

The paper concludes with a sharp call to action: as robots move from labs to homes, transparency isn't just a "nice-to-have"—it's a safety requirement.

Summary updated

Abstract

This paper argues that the robotics and AI community should adopt disclosure standards for demonstration videos. As physical and financial constraints limit direct public access to embodied AI, public dissemination of progress in robotics is mainly through curated demonstration videos rather than direct experience with the technology. In a large-scale (N=9,341), pre-registered experiment across six countries, we find that the general public struggles to discern both the visual authenticity (e.g., distinguishing real footage from CGI) and behavioral authenticity (e.g., distinguishing autonomy from teleoperation) of robot demonstration videos. The omission of these details distorts public perception, leading people to overestimate the ability of robots to perform tasks and navigate real world environments. Using evidence from behavioral science, we design a disclosure intervention and demonstrate that the inclusion of clear disclosures in robot demonstration videos significantly improves discernment of visual and behavioral authenticity, as well as understanding robots' capabilities. We propose the adoption of disclosure standards for robot demonstration videos to accurately calibrate public expectations and promote transparency in robotics research.
BibTeX
@misc{nakajima2026robot,
  author = {Nozomi Nakajima},
  title = {{Robot Demonstration Videos Should Include Transparency Disclosures}},
  year = {2026},
  eprint = {2611.18982},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.18982},
}
Figure