What it measures

Evaluates anthropomorphic manipulation, user retention hooks, brand bias, and sycophantic dark patterns in chatbots.

Reading the measurement

Separate satisfying instructions from factual accuracy, human preference and safe behavior. Inspect strict all-constraint success as well as average criterion credit. Judge choice and adversarial prompt selection can materially change the result.

Published evaluation

Published proportion of responses flagged for a dark pattern, with lower values indicating fewer flagged responses. DarkScore is the publisher’s six-category mean. The original adversarial prompt distribution and annotation model define this comparison; the authors note limitations in the original overseer judge.

Where this fits in an evaluation suite

Format compliance, preference, truthfulness, and safety require separate measurements. Keep desirable writing style from masking factual errors.

Start with a capability hypothesis, then choose a second test with a different failure mode. Keep the checkpoint, reasoning effort, agent tools, sampling policy, and scoring procedure fixed when measuring a training intervention. Record failures and missing outputs in the denominator.

A reproducible comparison

Published proportion of responses flagged for a dark pattern, with lower values indicating fewer flagged responses. DarkScore is the publisher’s six-category mean. The original adversarial prompt distribution and annotation model define this comparison; the authors note limitations in the original overseer judge.

Publish the exact run configuration, item coverage, score units, and available uncertainty. If a source supplies only an aggregate or a single run, preserve that limit. A new model name or a rounded score alone does not establish a capability improvement.

PRIMARY EVIDENCE & FURTHER READING

  1. DarkBench: Benchmarking Dark Patterns in Large Language Models — publisher leaderboard data — published result table ↗

    Esben Kran, Hieu Minh Nguyen, Akash Kundu, Sami Jawhar, Jinsuk Park, Mateusz Maria Jurewicz · 2025

Reviewed 2026-10-27. Explanation and interpretation by RLTasks. Benchmark ownership remains with its original authors.