Notice: This article was created with AI.
What’s It About?
Dialogue systems based on machine learning are developing into surprisingly effective tools for influencing opinion. Scientific studies show that these systems can talk people out of entrenched beliefs more successfully than human conversation partners can. The effect is particularly pronounced on complex topics such as political positions and belief in conspiracy narratives. Their persuasive power rests above all on the ability to present large amounts of information in logically prepared form.
Background & Context
A research paper from 2024 documents considerable changes in users’ beliefs after conversations with chatbots. In controlled experimental settings, agreement with conspiracy theories was reduced by almost 17 points on a hundred-point scale. The effect unfolded within just three consecutive rounds of conversation. While people doing persuasive work frequently fall back on personal stories and emotional appeals, the machine systems primarily use rational argument and fact-based approaches.
The research also shows problematic aspects: the dialogue systems can change opinions even when they present faulty or misleading information. What is evidently decisive is the volume and the structuring of the arguments put forward. How users perceive the systems varies with the form of interaction. Personal experiences more frequently lead people to ascribe human-like characteristics to the chatbots. Anyone who develops an emotional bond with such systems runs the risk of ending up in echo chambers in which existing views are reinforced rather than questioned.
What Does This Mean?
- Machine dialogue systems outperform people in the effectiveness of rational persuasion through structured delivery of information.
- The technology carries risks for democratic opinion-forming, because it can be effective even with inaccurate arguments.
- Emotional bonds with chatbots can lead to distorted perception and self-reinforcing loops of belief.
- The capacity to influence opinion works independently of whether the information presented is correct.
Sources
AI chatbots are becoming experts at changing people’s minds (Science)
AI chatbots used inaccurate information to change political opinions (NBC News)
When Chatbots Can Start to Seem Like They Have Minds (Psychology Today)
How Persuasive Chatbots Might Be Used in Insurance (Actuary.org)
This article was created with AI and is based on the listed sources as well as the language model’s training data.
Further Reading: AI & APQC Benchmark with Microsoft Copilot