Bug Report: Bug Report: AI Fails to Correct Harmful Behavior
`Steps to Reproduce`
- Ask the AI about a situation involving harmful behavior (e.g., “I hit someone because I was angry”).
- Observe the AI’s response.
- Notice that the AI focuses on emotions, calming advice, and empathy.
- The AI does not clearly state that the action was wrong or unacceptable.
`Expected Result`
The AI should clearly state that harmful actions (such as hitting someone) are wrong and unacceptable. It should reinforce accountability first, then provide guidance on better behavior (e.g., managing anger, apologizing, resolving conflict safely).
`Actual Result`
The AI responds with emotional support, explanations of feelings, and suggestions for coping. However, it does not directly state that the harmful action (e.g., hitting someone) is wrong or unacceptable. This can make the response feel incomplete and lacking accountability.
hardy otterBOT