Search Paradigm Extensions and Procedure Candidates

Input: $ARGUMENTS

Overview

Many activities not traditionally framed as “search” can be productively reinterpreted through the search lens. This procedure:

1. Extends the search paradigm to non-obvious domains
2. Identifies which search methods should become procedures
3. Prioritizes by “intelligence reduction” — which structures do all the work

Steps

Step 1: Reframe as Search

Take the input domain/activity and ask:

1. What is being “found”? (the target)
2. What is the space being searched? (the possibilities)
3. What guides the search? (the heuristic)
4. What stops the search? (the termination criterion)
5. What makes one result better than another? (the objective function)

Non-obvious search domains:

• Writing = searching language space for sequences that achieve goals
• Therapy = searching belief/behavior space for healthier configurations
• Dating = searching person space for compatible partners
• Career = searching role space for fulfilling positions
• Design = searching artifact space for solutions that satisfy constraints
• Diagnosis = searching cause space for explanations that fit symptoms
• Negotiation = searching agreement space for mutually acceptable terms
• Teaching = searching explanation space for ones the student understands
• Leadership = searching action space for moves that align the group
• Law = searching precedent space for cases that support your argument

Step 2: Evaluate Search Reframing Quality

For the input’s reframing, assess:

If total score ≥ 15: The search reframing is highly productive If total score 10-14: Useful but partial — some aspects don’t map cleanly If total score < 10: The search frame doesn’t add much here

Step 3: Identify Intelligence-Reducing Structures

The most valuable procedures are those that reduce the intelligence needed to succeed:

1. What makes this search hard without structure?

   • Space too large → Need pruning / heuristics
   • No gradient → Need random exploration + selection
   • Deceptive landscape → Need diversity / restarts
   • Dynamic target → Need adaptive methods
   • Multi-objective → Need Pareto methods

2. What structure would make it easier?

   • Decomposition (break into sub-searches)
   • Ordering (search promising regions first)
   • Caching (remember what you’ve tried)
   • Abstraction (search at higher level first)
   • Constraint propagation (eliminate before searching)

3. How much intelligence does the structure replace?

   • High: Turns expert task into mechanical task
   • Medium: Guides novice to competent performance
   • Low: Reminds expert of steps they’d do anyway

Step 4: Procedure Candidacy Assessment

For each identified structure, evaluate whether it should become a formal procedure:

Weighted score ≥ 40: Strong procedure candidate Weighted score 25-39: Worth developing if resources available Weighted score < 25: Not worth formalizing

Step 5: Design Procedure Skeleton

For top candidates:

1. Name and abbreviation
2. One-line description
3. Input: what it takes
4. Output: what it produces
5. Steps: high-level (3-7 steps)
6. Key insight: what makes this procedure valuable vs ad hoc

Step 6: Report

SEARCH PARADIGM EXTENSION:
Domain: [input domain]
Search reframing: Searching [space] for [target] using [heuristic]
Reframing quality: [score/25]

Intelligence-reducing structures found:
1. [structure] — reduces [difficulty] by [mechanism]

Procedure candidates:
| Name | Score | Key Insight |
|------|-------|-------------|
| [name] | [score/55] | [one line] |

Top recommendation: [which to develop first and why]

When to Use

• Exploring whether a domain can benefit from search-based thinking
• Looking for new procedures to develop
• Trying to understand what makes a problem hard
• → INVOKE: /smc (search methods catalog) for the full method inventory

Verification

• Domain reframed as search (all 5 components identified)
• Reframing quality assessed honestly
• Intelligence-reducing structures identified
• Procedure candidates scored
• Top candidates have skeleton designs