# Personal Decision Ladder starter kit

Use this kit inside the project you created in **Build a Project That Remembers**. Keep sensitive or confidential information out of any AI tool unless you have permission to use it there.

AI can help clarify, research, organize, calculate, compare, and challenge. The final choice is mine.

## 1. Choose a safe decision

Use a real decision only when it is reversible, non-urgent, and low-to-medium stakes. For urgent, irreversible, safety-critical, medical, legal, or investment decisions, choose a fallback decision for this exercise instead.

### Decision brief

- Decision owner:
- Decision question:
- Desired outcome:
- Options:
- Deadline:
- Constraints:
- Non-negotiables:
- How reversible is this decision?
- Outside this decision:

## 2. `memory/DECISION-METHOD.md`

```md
# Decision Method

## Ownership

I make the final decision. AI supports the reasoning and does not choose for me.

## Suitable decisions

- Reversible:
- Low-to-medium stakes:
- Enough time to check material information:

## Decisions needing different support

- Urgent or safety-critical:
- Medical, legal, or investment:
- Irreversible or outside my authority:

## Values and non-negotiables

- Values I want to protect:
- Outcomes I will not accept:

## Risk and reversibility

- My tolerance for downside:
- What makes an option safely reversible:
- When I prefer a small test before a larger commitment:

## Evidence standards

- Information I will verify:
- Source types I prefer:
- Uncertainty that must remain visible:

## Assumption checks

- What might I be taking for granted?
- What evidence would change my view?
- Which option am I emotionally favoring?
- Am I comparing the same timeframe and scope?

## AI support contract

AI may clarify, research, label, calculate, compare, challenge, and draft.
AI may not choose my values, set or change weights without approval, invent evidence, hide uncertainty, recommend an option, or save unapproved memory.

## Memory approval

Every durable update follows:
AI drafts -> I check meaning and evidence -> I correct or approve -> Codex updates -> I confirm the visible result.

## Method revision history

### YYYY-MM-DD

- Evidence from decision:
- Reusable change:
- Why it should apply again:
```

## 3. Criteria and evidence plan

Choose three to seven criteria. Mark a criterion `Threshold` when an option must pass it regardless of the weighted score.

| Criterion | Why it matters | Weight | Threshold? | Evidence needed |
| --- | --- | ---: | --- | --- |
|  |  |  |  |  |
|  |  |  |  |  |
|  |  |  |  |  |

**Weights total: 100**

Use a 1–5 score only when the evidence supports it:

- `1` = clearly weak against this criterion
- `2` = somewhat weak
- `3` = mixed or acceptable
- `4` = strong
- `5` = clearly strong
- `?` = not enough information to score

## 4. Evidence table

| Option | Criterion | Type | Evidence or claim | Direct source | Relevant date | Checked | Confidence | Score |
| --- | --- | --- | --- | --- | --- | --- | --- | ---: |
|  |  | Fact |  |  |  |  |  |  |
|  |  | Inference |  |  |  |  |  |  |
|  |  | Assumption |  |  |  |  |  |  |
|  |  | Unknown |  |  |  |  |  | ? |

- **Fact:** Supported by learner-confirmed information or an opened source.
- **Inference:** A reasoned interpretation based on identified facts.
- **Assumption:** A belief being used without enough confirmation.
- **Unknown:** Information that remains unavailable or unresolved.

## 5. Comparison

For each supported score:

`weighted contribution = (score ÷ 5) × weight`

Add the weighted contributions for a provisional result from 0 to 100. Leave unknown criteria as `?`; do not make an incomplete option look precise.

| Option | Thresholds passed? | Known weighted result | Unknown criteria | Important trade-off |
| --- | --- | ---: | --- | --- |
|  |  |  |  |  |
|  |  |  |  |  |
|  |  |  |  |  |

### Sensitivity check

- Move 10 weight points between the two most influential criteria while keeping the total at 100.
- Move one uncertain score up or down by one point when that range is plausible.
- Record whether the current ranking changes.
- Record the missing fact most likely to change the comparison.

AI may report the current score produced by my approved inputs. That analytical result is not a recommendation.

## 6. Challenge prompts

Ask AI:

> Separate facts, inferences, assumptions, and unknowns in this comparison. Identify the weakest material evidence and one credible argument against the currently highest-scoring option. Show which reasonable weight or score change could alter the ranking. Do not recommend or choose an option.

Then answer:

- What may I be overlooking?
- What trade-off am I accepting?
- What evidence would make me reconsider?
- Does my intuition reveal a missing criterion, or am I avoiding an uncomfortable result?

## 7. `memory/DECISIONS.md`

```md
## YYYY-MM-DD — Decision title

- Owner:
- Status: Decided
- Decision question:
- Desired outcome:
- Options considered:
- Deadline:
- Constraints and non-negotiables:
- Reversibility:

### Criteria and weights

| Criterion | Weight | Threshold? |
| --- | ---: | --- |
|  |  |  |

### Evidence summary

- Facts:
- Inferences:
- Assumptions:
- Unknowns:
- Material sources and dates:

### Comparison

- Threshold result:
- Weighted result:
- Sensitivity:
- Strongest opposing view:

### My choice

- Final choice:
- Why I chose it:
- Trade-offs I accept:
- Alternatives rejected:
- Confidence:
- What could make me reconsider:

### Outcome review trigger

- Review date or observable event:
```

Before saving, say in your own words:

> The final choice is mine. The comparison informed my judgment but did not make the decision.

## 8. Outcome review

Return when the review date or observable event arrives.

Add this to `memory/LEARNINGS.md`:

```md
## YYYY-MM-DD — Learning from decision title

- Linked decision:
- Expected outcome:
- Actual outcome:
- What surprised me:
- Evidence that proved useful:
- Assumptions that held:
- Assumptions that failed:
- Reasoning behavior that helped:
- Reasoning behavior to change:
- Reusable learning:
- Should DECISION-METHOD.md change? Yes / No
- Approved method change and reason:
```

Do not change `memory/DECISION-METHOD.md` merely because the outcome was disappointing. Change it only when the outcome provides a lesson that should apply again.

## 9. Memory approval checklist

- [ ] I chose the criteria and weights.
- [ ] The weights total 100.
- [ ] I opened material external sources.
- [ ] Facts, inferences, assumptions, and unknowns are distinct.
- [ ] Missing or conflicting information remains visible.
- [ ] The calculation uses my approved inputs.
- [ ] AI did not turn the score into a recommendation.
- [ ] I wrote or approved the final choice and reasoning.
- [ ] The review trigger is observable.
- [ ] I inspected every proposed memory update before it was saved.

## 10. Fallback decisions

### A. Which skill should I learn next?

- Options: spreadsheet analysis, presentation design, or project documentation.
- Constraints: four hours per week, a modest course budget, and a result needed within one month.
- Sample evidence: course syllabus, total practice time, current work needs, and one small trial lesson.

### B. Which small project should I prioritize this month?

- Options: improve a recurring report, organize project memory, or create a simple personal profile page.
- Constraints: eight available hours and no confidential data.
- Sample evidence: people affected, hours saved, dependencies, and reversibility.

### C. Which workflow should I test for a recurring task?

- Options: improve the current manual checklist, use an existing approved tool, or run a one-week parallel trial.
- Constraints: no account purchase during the exercise and no confidential uploads.
- Sample evidence: current task time, error rate, official feature documentation, and a privacy-safe trial.

### D. Which non-urgent course, event, or travel option fits best?

- Options: choose two or three real or fictional alternatives.
- Constraints: fixed budget, available dates, travel time, and cancellation conditions.
- Sample evidence: official schedule, full cost, cancellation policy, and learner priorities.

### E. Should I continue, change, or stop a low-risk personal project?

- Options: continue unchanged, reduce scope for a short trial, or stop.
- Constraints: time available, original purpose, and commitments already made.
- Sample evidence: time spent, useful progress, remaining effort, and what has changed since starting.