The Placebo Effect: AI-Enhanced Belief Encourages Risk-Taking

summary: Belief in AI enhancements may lead to increased risk-taking behavior. In this study, participants were informed that AI applications improved cognitive performance during a virtual card game.

Despite the lack of real AI assistance, participants showed higher risk taking, indicating potential placebo effects in technology application.

The findings highlight the importance of evaluating real-world AI benefits and considering user expectations in the development process.

important facts:

1. The study included a virtual card game in which participants were told they would be assisted by an AI application, which increased risk-taking behavior without actually augmenting AI.
2. This result suggests that a placebo effect exists in the use of technological applications, similar to the placebo effect observed with pharmaceuticals.
3. Researchers emphasize the importance of evaluating the true benefits of AI applications before release and considering user expectations during development.

sauce: LMUs

Human augmentation technology refers to technical assistance that enhances human capabilities. This includes things like augmented reality headsets, as well as exoskeletons.

A study by LMU’s Human-Centric Ubiquitous Media Chair found that users have high expectations of the effectiveness of these technologies.

As soon as they believe AI is improving their cognitive abilities, they start taking risks. And it does this regardless of whether the AI ​​is actually assisting it or not.

“Hype about AI applications influences user expectations, which can lead to more risky behavior,” said postdoctoral fellow at the Chair of the Human-Centric Ubiquitous Media Committee and lead author of the study. says Steven Villa.

Exclusion of placebo effect

In this study, participants were informed that they would be assisted by an AI application that enhanced their cognitive abilities during a virtual card game.

There was actually no such AI enhancement. Nevertheless, participants were willing to take higher risks as soon as they believed they were benefiting from AI.

This study points out that in this kind of technical application there may be placebo effects similar to well-established placebo effects in pharmaceuticals.

“As people interact more and more with intelligent systems, it’s important to understand the potential placebo effect so that we can build systems that provide real support,” said Professor of Computer Science at the LMU. Albrecht Schmidt said.

Before releasing an AI application, researchers recommend evaluating its real-world benefits, given possible placebo effects. Additionally, they advise technology companies to involve users and their expectations more deeply in the development process.

About this AI and psychology research news

author: Constanze Dorero
sauce: LMUs
contact: Constanze Drulo – LMU
image: Image credited to Neuroscience News

Original research: open access.
“Human Augmentation Placebo Effect: Anticipation of Cognitive Augmentation Increases Risk-Taking BehaviorBy Stephen Villa et al. computer in human behavior

overview

Human Augmentation Placebo Effect: Anticipation of Cognitive Augmentation Increases Risk-Taking Behavior

Human augmentation technology uses technology to enhance human capabilities. This study explores the placebo effect of augmentation technology.

Thirty naïve participants were instructed to either augment with cognitive augmentation techniques or not use the augmentation system when performing the Columbia Card task. In this risk-taking method, participants flip over their winning and losing cards.

The fake augmentation system allegedly consisted of a brain-computer interface tuned to play inaudible sounds that enhance cognitive function. However, no sound was played under all conditions.

Using Bayesian statistical modeling, we demonstrate the placebo effect in human augmentation, where continued belief in improvement remains after using a pseudo-system and risk-taking increases contingent on heightened expectations.

In addition, we identify differences in EEG event-related potentials that occur during the sham condition when flipping loss cards.

Finally, we integrate our findings into the theory of human augmentation and discuss implications for future evaluations of augmentation techniques.