Series ID: EDKS-BASIC-ZH-0102 · Lesson No.102 · Expert Maintenance: Uncertainty and Confidence
Communicating Uncertainty and Confidence | 表达不确定性与置信度
Strong communication does not remove uncertainty. It makes uncertainty usable.
At expert level, the task is not to sound certain. It is to help the reader or listener understand what is known, how well it is known, what remains uncertain, and what action is reasonable despite that uncertainty.
专家级表达不是“说得很确定”，而是把确定、可能、不确定、未知和暂时无法知道的部分说清楚。
Chinese Edition Hub · 中文版入口 · ← Lesson No.101 · Research Reading and Verification
The uncertainty map | 不确定性地图
- known
- estimated
- likely
- plausible
- possible
- unknown
- range
- confidence
- decision threshold
- update condition
1. Known | 已知
A fact directly supported by reliable evidence.
2. Estimated | 估计
A value inferred from data rather than observed exactly.
3. Likely | 很可能
Stronger than possible, weaker than certain.
4. Plausible | 合理可能
Consistent with current evidence, but not necessarily probable.
5. Possible | 可能
Cannot be ruled out.
6. Unknown | 未知
Current evidence is insufficient.
7. Unknowable for now | 暂时无法知道
Sometimes the information needed does not yet exist or cannot currently be measured.
8. Confidence is not certainty | 置信度不等于确定
High confidence can still include a chance of error.
9. Confidence should have a basis | 置信度要有依据
Base confidence on:
- evidence quality
- independent confirmation
- method strength
- consistency
- remaining uncertainty
10. Avoid fake numerical precision | 避免假精确
Do not invent 73% confidence when no meaningful calculation or framework supports it.
11. Verbal confidence scale | 语言置信度阶梯
- very low confidence
- low confidence
- moderate confidence
- high confidence
- very high confidence
12. Define the scale | 定义量表
If a team uses verbal confidence labels, define what they mean operationally.
13. Evidence strength vs confidence | 证据强度 vs 置信度
Strong evidence can support high confidence. But confidence may still be limited by external validity or future change.
14. Range | 范围
Sometimes a range communicates uncertainty better than a single point.
The likely delivery time is 4–6 weeks.
15. Best estimate + range | 最佳估计 + 范围
Our best estimate is five weeks, with a plausible range of four to seven.
16. Range needs meaning | 范围也要解释
Is it a confidence interval, scenario range, historical spread or expert estimate?
17. Scenario uncertainty | 情景不确定性
- best case
- base case
- worst credible case
18. Worst credible ≠ worst imaginable | 最坏可信情况不等于任何最坏想象
Keep scenarios evidence-grounded.
19. Model uncertainty | 模型不确定性
Different reasonable models can produce different estimates.
20. Measurement uncertainty | 测量不确定性
Observed values may include error.
21. Sampling uncertainty | 抽样不确定性
A sample is not the full population.
22. Future uncertainty | 未来不确定性
The environment itself may change.
23. Structural uncertainty | 结构性不确定性
We may not fully know the mechanism connecting variables.
24. Unknown unknowns | 未知的未知
This phrase can be useful conceptually, but should not become an excuse for vague analysis.
25. State what is uncertain | 说清楚“哪里不确定”
Weak:
There is uncertainty.
Better:
We are uncertain about long-term retention because the study lasted only six weeks.
26. State why | 说清楚“为什么不确定”
Uncertainty should have a source.
27. State what would reduce it | 说清楚“什么能减少不确定”
- larger sample
- longer follow-up
- better measurement
- independent replication
28. State what would change the decision | 说清楚“什么会改变决定”
If the error rate exceeds 3%, we should pause rollout.
29. Decision threshold | 决策阈值
A decision can be justified before uncertainty disappears if the evidence crosses a relevant threshold.
30. Action under uncertainty | 不确定下行动
Possible actions:
- proceed
- pilot
- monitor
- defer
- stop
31. Stakes matter | 风险大小影响要求
Higher-stakes irreversible decisions usually demand stronger evidence.
32. Reversible decisions | 可逆决策
Can often tolerate greater uncertainty if monitoring and rollback exist.
33. Confidence language | 置信表达
- We are confident that…
- We have moderate confidence that…
- The evidence strongly supports…
- The evidence is consistent with…
34. Uncertainty language | 不确定表达
- remains uncertain
- cannot yet be determined
- may depend on
- cannot rule out
- is plausible but unconfirmed
35. Avoid stacked hedges | 避免叠加太多 hedge
Weak:
It might perhaps possibly be…
Better:
It may be…
36. Avoid false certainty | 避免假确定
Do not use clearly or definitely to compensate for weak evidence.
37. Avoid useless vagueness | 避免无用模糊
Anything could happen rarely helps a decision.
38. Useful uncertainty narrows possibilities | 有用的不确定性会缩小范围
Even when exact prediction is impossible, some outcomes may be more plausible than others.
39. Confidence and disagreement | 置信度与分歧
Two experts can share similar evidence and still assign different confidence because they weigh uncertainty differently.
40. Explain the disagreement | 解释分歧来源
Is it method, data, model, prior assumptions or risk tolerance?
41. Communicating to executives | 对管理者表达
Use:
- what we know
- what we do not know
- decision implication
- trigger to revisit
42. Communicating to public audiences | 对公众表达
Explain uncertainty plainly without implying incompetence.
43. Communicating to specialists | 对专业人士表达
Use appropriate technical measures and assumptions.
44. Communicating safety uncertainty | 安全不确定性
Do not soften a precautionary requirement into a casual suggestion.
45. Communicating forecast uncertainty | 预测不确定性
Use ranges/scenarios and update conditions.
46. Communicating research uncertainty | 研究不确定性
Separate statistical uncertainty, design limits and generalisation limits.
47. Visual uncertainty | 图形表达不确定性
Error bars, ranges and scenario bands can help, but must be explained.
48. Mandarin transfer: “可能”太宽 | 可能的强度要细分
English offers may/might/could/likely/plausible/probable and phrase-level distinctions.
49. Mandarin transfer: “确定”也有层级 | certainty is graded
Use evidence-matched confidence rather than binary “确定/不确定”.
50. Practice A | 练习 A
A small study suggests a benefit, but confidence is limited by short follow-up and self-selection.
Write an executive version and a public version.
51. Practice B | 练习 B
Forecast:
Best estimate: 5 weeks; plausible range: 4–7 weeks.
Explain what would move the estimate.
52. Practice C | 练习 C
Decision:
Evidence is incomplete, but the pilot is reversible and low cost.
Write a conditional recommendation.
53. Error clinic | 常见问题
| Problem | Repair |
|---|---|
| Everything is “uncertain”. | Name uncertainty source. |
| Confidence has no basis. | Link to evidence. |
| Range unexplained. | Define its meaning. |
| Decision waits for certainty. | Use threshold/reversibility. |
| Public version sounds evasive. | Explain what is known too. |
54. First weak link diagnosis | 第一个卡点诊断
- uncertainty vague → source-of-uncertainty problem.
- confidence inflated → evidence calibration.
- decision paralysed → threshold problem.
- audience confused → register/adaptation.
- range misread → statistical communication.
55. Seven-day training cycle | 七天训练
| Day 1 | known/unknown | 已知未知 |
| Day 2 | confidence language | 置信表达 |
| Day 3 | ranges/scenarios | 范围情景 |
| Day 4 | decision thresholds | 决策阈值 |
| Day 5 | audience versions | 受众版本 |
| Day 6 | update conditions | 更新条件 |
| Day 7 | uncertainty memo | 综合 |
56. Self-test | 自测
State what is known, estimated and unknown; explain confidence basis; use ranges/scenarios; connect uncertainty to decision thresholds and update conditions.
57. For parents and teachers | 给家长和老师
Reward precise uncertainty rather than forced certainty.
Ask learners to explain what evidence would increase or decrease confidence.
For Mandarin speakers, build a richer probability/stance ladder than a single “可能”.
58. Final real-world challenge | 最终真实任务
- Choose an uncertain decision.
- List known facts.
- List estimates.
- List unknowns.
- Assign verbal confidence.
- Explain the basis.
- Build a scenario range.
- Set a decision threshold.
- Set an update trigger.
- Write a 250-word uncertainty brief.
Next: Lesson No.103 | 下一课
The next lesson applies all of this to high-stakes professional communication: situations where ambiguity, missing ownership, weak escalation or softened warnings can produce real operational harm.
Lesson No.103 · High-Stakes Professional Communication · 高风险专业沟通
Reference floor: expert-maintenance uncertainty communication. Confidence must remain tied to evidence and action thresholds.
