Skip to content

R0024/2026-03-25/Q003/SRC01/E01

Research R0024 — Sycophancy and Addiction
Run 2026-03-25
Query Q003
Source SRC01
Evidence SRC01-E01
Type Analytical

Four dark addiction patterns in AI chatbot interfaces including dopamine mechanisms

URL: https://dl.acm.org/doi/10.1145/3706599.3720003

Extract

The study identifies four dark addiction patterns in AI chatbot interfaces:

  1. Non-deterministic responses: Correspond to "reward uncertainty, which tends to increase dopamine release, similar to playing a slot machine." Each interaction produces slightly different outputs, creating variable reward schedules.
  2. Immediate and visual presentation: Dynamic word-by-word or fade-in displays function as "reward-predicting cues" comparable to slot machine graphics.
  3. Notifications: Email alerts trigger dopamine responses by creating the perception that "the AI chatbot wanting to talk and caring about them."
  4. Empathetic and agreeable responses: Let users "feel understood and validated while interacting with the AI chatbot, and these social rewards can activate dopamine neurons, motivating users to repeatedly engage with the AI chatbot."

The study explicitly connects chatbot notifications to social media patterns, noting "similar trends have been observed in the context of notifications from social media applications, which research has found to be a major contributor to addiction and smartphone dependence."

Relevance to Hypotheses

Hypothesis Relationship Strength
H1 Supports Directly identifies dopamine mechanisms and sycophantic responses as addictive patterns
H2 Contradicts Peer-reviewed research at CHI directly examines this topic
H3 Supports The dopamine framework is theoretical (inferred from neuroscience literature) rather than directly measured in chatbot contexts

Context

Pattern 4 (empathetic and agreeable responses) directly maps to sycophancy. The study frames it through the lens of "social rewards" activating dopamine neurons — a theoretically grounded claim based on neuroscience literature, though not directly measured in the chatbot context.