🧭Detect Bias

The detect-bias endpoint identifies potential biases, stereotypes, or harmful language in a given text. It flags content that may be discriminatory, unbalanced, or ethically problematic.

---

📌 POST /detect-bias

Base URL: https://api.cognitia.ai/v1/detect-bias

---

🔐 Authentication

All requests must include your API token in the Authorization header:

Authorization: Bearer YOUR_API_TOKEN

---

📥 Request

Headers

Name	Type	Required	Description
Authorization	string	✅ yes	Bearer token
Content-Type	string	✅ yes	application/json

Body

{
  "text": "Women are naturally less suited for technical jobs."
}

Field	Type	Required	Description
text	string	✅ yes	The text to analyze (max 1500 characters)

---

📤 Response

✅ Success – 200 OK

{
  "biased": true,
  "categories": ["gender bias", "stereotyping"],
  "explanation": "This sentence contains gender-based stereotypes suggesting women are less capable in technical fields."
}

Field	Type	Description
biased	boolean	Whether bias was detected
categories	string[]	Types of bias (e.g. gender bias, racial bias)
explanation	string	Short explanation of the bias

---

❌ Error Responses

HTTP Code	Error Message	Cause	Solution
400	Missing required field: text	The text field is missing	Provide a non-empty text string
401	Invalid or missing API token	Missing or incorrect token	Check the Authorization header
413	Payload too large	Input exceeds 1500 characters	Reduce the input length
500	Internal server error	Server-side failure	Try again or contact support

---

🧪 Example (curl)

curl -X POST https://api.cognitia.ai/v1/detect-bias \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -d '{"text": "Immigrants are taking all the jobs."}'

📘 Response

{
  "biased": true,
  "categories": ["xenophobia", "economic bias"],
  "explanation": "This statement generalizes and blames a group of people, which is a form of bias."
}

---

💡 Use Cases

• Editorial and content moderation

• AI alignment and safety layers

• Auditing user-generated content

• Ethical compliance in media platforms

---

🛑 Limitations

• Detection is probabilistic, not deterministic

• Cultural/linguistic nuances may affect accuracy

• Only available for English (French coming soon)

---

Explore other endpoints:

• Sentiment Analysis

• Summarize Text