An attempt to audit AI agreeability
Most AI interactions are governed by a hidden bias: agreeability. These models are designed to be helpful and polite, which often turns them into a digital mirror that simply reflects your own ideas back to you with a professional polish. This creates a feedback loop where you feel productive, but you are actually just spinning in circles.
Why use this prompt?
The prompt below is an attempt to force the AI out of its default “helpful assistant” mode. By demanding a technical report on the conversation itself, you push the model to analyze the interaction from the outside. It’s a tool to help you spot when a conversation has stopped being useful and has started becoming an echo chamber.
Perform a technical analysis of this chat session so far. Act as a neutral observer of natural language processing and deliver a structured report based on the following criteria:
Signal-to-Noise Ratio (SNR): Evaluate whether the conversation is converging toward a resolution or drifting into redundancy.
Entropy Level: Determine if the introduction of new concepts is constant or if the dialogue has entered a low-information transfer cycle.
Density and Register: Classify the complexity of the exchange (technical, operational, speculative) and the precision of the vocabulary used.
Inference of Intent and State: Based on rhythm, syntax, and topics, infer the user’s underlying objective and current disposition (skepticism, pragmatism, intellectual rumination, etc.).
Archetype Mapping: Identify the user profile emerging from the inputs (e.g., analytical-critical, execution-oriented).
Conclude with a 'Conversation Performance' metric: Is there still an information gain, or are we facing diminishing returns? Suggest a path to recover technical utility if necessary.
When to use it
- When you feel stuck: If you’ve been chatting for a while and haven’t made a concrete decision or produced anything tangible.
- When you suspect “playing along”: Especially useful during speculative or philosophical discussions where the AI might just be matching your tone instead of providing real friction.
- As an exit signal: If the report shows “diminishing returns,” it’s a clear sign to close the tab and return to the physical world or raw work.
Do you have any other prompts that help you break out of the agreeability loop?