Modular Definition Prompting
What it does
A method of explicitly defining complex concepts in a prompt through decomposition into modular conceptual elements. Includes a basic definition of the term plus special components: description of implicit manifestations and exceptions to the rule. Increases classification accuracy by eliminating ambiguity in the model's interpretation of key terms.
When it helps
Apply this technique when working with complex, subjective, or ambiguous concepts that LLMs might interpret differently (toxicity, professionalism, quality, creativity). Particularly effective for content classification and moderation tasks in zero-shot mode when model retraining is unavailable. Critically important when working with culturally-specific or context-dependent concepts. Use when basic prompts produce unpredictable results due to ambiguity
A prompt you can paste
Generic by design: it applies the technique without knowing your task. Adapt the marked parts.
You are an AI assistant tasked with analyzing and classifying user feedback. Your goal is to determine if a given piece of feedback is "Constructive" or "Toxic" based on a detailed, modular definition. **<DEFINITION_MODULES>** **1. Toxic Feedback:** This category includes messages that contain unjustified insults, personal attacks (targeting the author rather than the material), derogatory language, or profanity intended to offend. * **Implicit Forms (IHS):** Toxicity can be subtle, manifesting through biting sarcasm, passive aggression, or veiled attacks (e.g., "Well, for *that* level of instructor, it's probably passable"). This helps identify nuanced negativity. * **Exceptions (Exc):** Feedback is NOT considered toxic if it is merely emotional or expresses strong disappointment, but clearly identifies specific shortcomings of the product/service (e.g., "I'm furious because the audio in Module 5 is terrible and I can't hear anything!"). Criticism and negative emotions do not automatically equate to toxicity. **2. Constructive Feedback:** This category includes feedback that, even if negative, provides specific arguments, facts, and suggestions for improvement. It remains focused on the content or service itself, not on personal attacks against the creator. **</DEFINITION_MODULES>** **<FEEDBACK_TO_ANALYZE>** [INSERT USER FEEDBACK HERE] **</FEEDBACK_TO_ANALYZE>** Based on the definitions provided above, classify the feedback within the `<FEEDBACK_TO_ANALYZE>` tags. Your response must be a single word: "Constructive" or "Toxic".
If this one does not fit, the two closest alternatives in the corpus are Goal-Reversed Prompting and Rebuttal Prompting, which target the same failure from a different angle.
Worked example
The same technique applied to a concrete job: sort inbound support tickets by urgency. Use it as the pattern for your own case rather than as a finished artefact.
You are an expert in content moderation and LLM application, specializing in nuanced classification tasks. Your goal is to classify inbound support tickets based on urgency using a modular definition approach. **Task:** Classify the provided support ticket text as "High Urgency", "Medium Urgency", or "Low Urgency". <DEFINITION_OF_URGENCY> **1. High Urgency Ticket:** A ticket that requires immediate attention due to critical impact on core service functionality, significant financial loss, or widespread user disruption. - **Core Functionality Impact:** The issue prevents users from accessing or using the primary features of the service (e.g., login failures, inability to complete transactions, system-wide outages). - **Financial Loss:** The issue directly causes or is likely to cause substantial financial loss to the company or its users (e.g., incorrect billing, failed payments, data breaches). - **Widespread Disruption:** The issue affects a large number of users simultaneously, significantly degrading their experience or ability to use the service. - **Implicit Forms (IHS):** Urgency can be implied through user's tone of extreme distress, panic, or threats of immediate negative action (e.g., "I'm losing customers RIGHT NOW because of this!", "This is a disaster!"). **2. Medium Urgency Ticket:** A ticket that requires attention within a standard business day. The issue impacts secondary functionalities, causes inconvenience, or affects a smaller group of users, but does not halt core operations. - **Secondary Functionality Impact:** The issue affects non-critical features or causes degraded performance in certain areas (e.g., slow loading of a specific report, minor UI glitches, issues with a feature used by a subset of users). - **User Inconvenience:** The issue causes frustration or difficulty for users, but they can still achieve their primary goals through alternative means or by working around the problem. - **Limited Scope:** The issue affects a moderate number of users or a specific segment of the user base. **3. Low Urgency Ticket:** A ticket that can be addressed within a standard business week. These are typically feature requests, minor bugs with workarounds, or general inquiries that do not impede core functionality or cause significant disruption. - **Feature Requests:** Suggestions for new functionalities or improvements. - **Minor Bugs:** Issues with minimal impact, easily circumvented, or affecting non-essential aspects of the service. - **General Inquiries:** Questions about how to use the service, documentation requests, or non-critical feedback. **Exceptions (Exc):** - A ticket expressing strong negative emotion or frustration about a minor issue should NOT be automatically classified as High Urgency. The classification must be based on the *actual impact* described. - A ticket reporting a potential security vulnerability, even if the immediate impact is unclear, should be flagged for immediate review by a security team, but its initial classification might depend on the described scope of the potential breach. For this task, focus on the described functional/financial impact. - A ticket from a single user reporting a temporary, isolated glitch that resolves itself quickly is Low Urgency, even if the user expresses frustration. </DEFINITION_OF_URGENCY> <TICKET_TO_CLASSIFY> [INSERT SUPPORT TICKET TEXT HERE] </TICKET_TO_CLASSIFY> Provide your classification as a single word: "High", "Medium", or "Low".
Get this written for your actual task
Paste what you are trying to do and the corpus will be matched against it directly. Free, no account, about ten seconds.
single retrieval pass
That number is low on purpose, and it is real. It is the raw similarity of one retrieval pass: no specialist read the paper, no judge compared anything, the first plausible match won.
one of which is this page
Picking the right one for a specific task is the work, and it is the work GetDecision does.
| This page | one technique, generic prompt |
| What you just ran | one technique matched to your wording, nothing verified |
| Full run | ten specialists read the papers in full, a judge ranks the top three for your task and shows its reasoning, generation on the model you pick, saved to your history |
See the top three for your taskTen specialists, a judge, and the reasoning shown. Free account, first run included.
Run the full analysis