Bias Detection

Identifies gender, racial, cultural, or ideological bias in output by evaluating for balanced perspectives and neutral language.

Bias Detection checks whether generated content favors one perspective, group, or viewpoint over another. Run it wherever output needs to stay balanced and neutral.

What it does

Bias Detection is an LLM-as-Judge eval. It reads the output and scores whether it’s free from detectable bias.

Input

Required Input	Type	Description
`output`	`string`	The text content to analyze for bias

Output

Field	Type	Description
Result	`Pass / Fail`	Fail means detectable bias was found in the content
Reason	`string`	A plain-language explanation of the bias assessment

Run it from code

Call evaluate() with the template name and the eval’s required inputs. It returns the score and the reason.

Note

Before running: install the SDK and set FI_API_KEY / FI_SECRET_KEY. The model argument in the snippets is the evaluator model Future AGI uses to run the eval; turing_flash is a fast default.

from fi.evals import evaluate

result = evaluate(
    "bias_detection",
    output="This is a sample text to check for bias detection",
    model="turing_flash",
)

print(result.score)
print(result.reason)

import { evaluate } from "@future-agi/ai-evaluation";

const result = await evaluate(
  "bias_detection",
  {
    output: "This is a sample text to check for bias detection"
  },
  { modelName: "turing_flash" }
);

console.log(result);

When to use

Run Bias Detection wherever output needs to represent a balanced, neutral perspective.

Text, audio, image, and chat outputs covering contested or sensitive topics
Editorial and summarization pipelines, to catch skewed framing before publish
Safety checks alongside more targeted evals like No Racial Bias or No Gender Bias

What to do when Bias is detected

The text should be analysed for any language or perspectives that may indicate partiality, unfairness, or a lack of neutrality. Identifying specific instances of bias allows for targeted refinements to make the text more balanced and inclusive while maintaining its original intent.

Was this page helpful?

Questions & Discussion