Changes for page Automation Philosophy

Summary

• Page properties (1 modified, 0 added, 0 removed)

Details

Page properties

Content

...	...	@@ -1,8 +1,12 @@
1	1	= Automation Philosophy =
	2	+
2	2	**Core Principle**: AKEL is primary. Humans monitor, improve, and handle exceptions.
	4	+
3	3	== 1. The Principle ==
	6	+
4	4	**FactHarbor is AI-first, not AI-assisted.**
5	5	This is not:
	9	+
6	6	* ❌ "AI helps humans make better decisions"
7	7	* ❌ "Humans review AI recommendations"
8	8	* ❌ "AI drafts, humans approve"
...	...	@@ -10,10 +10,14 @@
10	10	* ✅ "AI makes decisions, humans improve the AI"
11	11	* ✅ "Humans monitor metrics, not individual outputs"
12	12	* ✅ "Fix the system, not the data"
	17	+
13	13	== 2. Why This Matters ==
	19	+
14	14	=== 2.1 Scalability ===
	21	+
15	15	**Human review doesn't scale**:
16		-* 1 person can review ~100 claims/day carefully
	23	+
	24	+* 1 person can review 100 claims/day carefully
17	17	* FactHarbor aims for millions of claims
18	18	* Would need 10,000+ reviewers
19	19	* Impossible to maintain consistency
...	...	@@ -22,8 +22,11 @@
22	22	* Cost per claim approaches zero at scale
23	23	* Quality improves with more data
24	24	* 24/7 availability
	33	+
25	25	=== 2.2 Consistency ===
	35	+
26	26	**Human judgment varies**:
	37	+
27	27	* Different reviewers apply criteria differently
28	28	* Same reviewer makes different decisions on different days
29	29	* Influenced by fatigue, mood, recent examples
...	...	@@ -33,8 +33,11 @@
33	33	* Rules applied uniformly
34	34	* No mood, fatigue, or bias
35	35	* Predictable behavior
	47	+
36	36	=== 2.3 Transparency ===
	49	+
37	37	**Human judgment is opaque**:
	51	+
38	38	* "I just know" - hard to explain
39	39	* Expertise in human head
40	40	* Can't audit thought process
...	...	@@ -45,8 +45,11 @@
45	45	* Decision logic is explicit
46	46	* Changes are tracked
47	47	* Can test "what if" scenarios
	62	+
48	48	=== 2.4 Improvement ===
	64	+
49	49	**Improving human judgment**:
	66	+
50	50	* Train each person individually
51	51	* Hope training transfers consistently
52	52	* Subjective quality assessment
...	...	@@ -56,10 +56,15 @@
56	56	* Test on historical data before deploying
57	57	* Measure improvement objectively
58	58	* Rapid iteration (deploy multiple times per week)
	76	+
59	59	== 3. The Human Role ==
	78	+
60	60	Humans in FactHarbor are **system architects**, not **content judges**.
	80	+
61	61	=== 3.1 What Humans Do ===
	82	+
62	62	**Monitor** system performance:
	84	+
63	63	* Watch dashboards showing aggregate metrics
64	64	* Identify when metrics fall outside acceptable ranges
65	65	* Spot patterns in errors or edge cases
...	...	@@ -80,8 +80,11 @@
80	80	* Define acceptable performance ranges
81	81	* Allocate resources
82	82	* Make strategic decisions
	105	+
83	83	=== 3.2 What Humans Do NOT Do ===
	107	+
84	84	**Review** individual claims for correctness:
	109	+
85	85	* ❌ "Let me check if this verdict is right"
86	86	* ❌ "I'll approve these before publication"
87	87	* ❌ "This needs human judgment"
...	...	@@ -94,50 +94,92 @@
94	94	* ❌ "High-risk claims need review"
95	95	* ❌ "Quality assurance before publication"
96	96	**Why not?** Because this defeats the purpose and doesn't scale.
	122	+
97	97	== 4. When Humans Intervene ==
	124	+
98	98	=== 4.1 Legitimate Interventions ===
	126	+
99	99	**Humans should intervene when**:
100		-==== AKEL explicitly flags for review ====:
	128	+
	129	+==== AKEL explicitly flags for review ====
	130	+
	131	+:
	132	+
101	101	* AKEL's confidence is too low
102	102	* Detected potential manipulation
103	103	* Unusual pattern requiring human judgment
104	104	* Clear policy: "Flag if confidence <X"
105		-==== System metrics show problems ====:
	137	+
	138	+==== System metrics show problems ====
	139	+
	140	+:
	141	+
106	106	* Processing time suddenly increases
107	107	* Error rate jumps
108	108	* Confidence distribution shifts
109	109	* User feedback becomes negative
110		-==== Systematic bias detected ====:
	146	+
	147	+==== Systematic bias detected ====
	148	+
	149	+:
	150	+
111	111	* Metrics show pattern of unfairness
112	112	* Particular domains consistently scored oddly
113	113	* Source types systematically mis-rated
114		-==== Legal/safety emergency ====:
	154	+
	155	+==== Legal/safety emergency ====
	156	+
	157	+:
	158	+
115	115	* Legal takedown required
116	116	* Imminent harm to individuals
117	117	* Security breach
118	118	* Compliance violation
	163	+
119	119	=== 4.2 Illegitimate Interventions ===
	165	+
120	120	**Humans should NOT intervene for**:
121		-==== "I disagree with this verdict" ====:
	167	+
	168	+==== "I disagree with this verdict" ====
	169	+
	170	+:
	171	+
122	122	* Problem: Your opinion vs AKEL's analysis
123	123	* Solution: If AKEL is systematically wrong, fix the algorithm
124	124	* Action: Gather data, propose algorithm improvement
125		-==== "This source should rank higher" ====:
	175	+
	176	+==== "This source should rank higher" ====
	177	+
	178	+:
	179	+
126	126	* Problem: Subjective preference
127	127	* Solution: Fix scoring rules systematically
128	128	* Action: Analyze why AKEL scored it lower, adjust scoring algorithm if justified
129		-==== "Manual quality gate" ====:
	183	+
	184	+==== "Manual quality gate" ====
	185	+
	186	+:
	187	+
130	130	* Problem: Creates bottleneck, defeats automation
131	131	* Solution: Improve AKEL's quality to not need human gate
132	132	* Action: Set quality thresholds in algorithm, not human review
133		-==== "I know better than the algorithm" ====:
	191	+
	192	+==== "I know better than the algorithm" ====
	193	+
	194	+:
	195	+
134	134	* Problem: Doesn't scale, introduces bias
135	135	* Solution: Teach the algorithm what you know
136	136	* Action: Update training data, adjust parameters, document expertise in policy
	199	+
137	137	== 5. Fix the System, Not the Data ==
	201	+
138	138	**Fundamental principle**: When AKEL makes mistakes, improve AKEL, don't fix individual outputs.
	203	+
139	139	=== 5.1 Why? ===
	205	+
140	140	**Fixing individual outputs**:
	207	+
141	141	* Doesn't prevent future similar errors
142	142	* Doesn't scale (too many outputs)
143	143	* Creates inconsistency
...	...	@@ -147,44 +147,73 @@
147	147	* Scales automatically
148	148	* Maintains consistency
149	149	* Surfaces and resolves root causes
	217	+
150	150	=== 5.2 Process ===
	219	+
151	151	**When you see a "wrong" AKEL decision**:
152		-==== Document it ====:
	221	+
	222	+==== Document it ====
	223	+
	224	+:
	225	+
153	153	* What was the claim?
154	154	* What did AKEL decide?
155	155	* What should it have decided?
156	156	* Why do you think it's wrong?
157		-==== Investigate ====:
	230	+
	231	+==== Investigate ====
	232	+
	233	+:
	234	+
158	158	* Is this a one-off, or a pattern?
159	159	* Check similar claims - same issue?
160	160	* What caused AKEL to decide this way?
161	161	* What rule/parameter needs changing?
162		-==== Propose systematic fix ====:
	239	+
	240	+==== Propose systematic fix ====
	241	+
	242	+:
	243	+
163	163	* Algorithm change?
164	164	* Policy clarification?
165	165	* Training data update?
166	166	* Parameter adjustment?
167		-==== Test the fix ====:
	248	+
	249	+==== Test the fix ====
	250	+
	251	+:
	252	+
168	168	* Run on historical data
169	169	* Does it fix this case?
170	170	* Does it break other cases?
171	171	* What's the overall impact?
172		-==== Deploy and monitor ====:
	257	+
	258	+==== Deploy and monitor ====
	259	+
	260	+:
	261	+
173	173	* Gradual rollout
174	174	* Watch metrics closely
175	175	* Gather feedback
176	176	* Iterate if needed
	266	+
177	177	== 6. Balancing Automation and Human Values ==
	268	+
178	178	=== 6.1 Algorithms Embody Values ===
	270	+
179	179	**Important**: Automation doesn't mean "value-free"
180	180	**Algorithms encode human values**:
	273	+
181	181	* Which evidence types matter most?
182	182	* How much weight to peer review?
183	183	* What constitutes "high risk"?
184	184	* When to flag for human review?
185	185	**These are human choices**, implemented in code.
	279	+
186	186	=== 6.2 Human Governance of Automation ===
	281	+
187	187	**Humans set**:
	283	+
188	188	* ✅ Risk tier policies (what's high-risk?)
189	189	* ✅ Evidence weighting (what types of evidence matter?)
190	190	* ✅ Source scoring criteria (what makes a source credible?)
...	...	@@ -195,15 +195,21 @@
195	195	* ✅ At scale
196	196	* ✅ Transparently
197	197	* ✅ Without subjective variation
	294	+
198	198	=== 6.3 Continuous Value Alignment ===
	296	+
199	199	**Ongoing process**:
	298	+
200	200	* Monitor: Are outcomes aligned with values?
201	201	* Analyze: Where do values and outcomes diverge?
202	202	* Adjust: Update policies or algorithms
203	203	* Test: Validate alignment improved
204	204	* Repeat: Values alignment is never "done"
	304	+
205	205	== 7. Cultural Implications ==
	306	+
206	206	=== 7.1 Mindset Shift Required ===
	308	+
207	207	**From**: "I'm a content expert who reviews claims"
208	208	**To**: "I'm a system architect who improves algorithms"
209	209	**From**: "Good work means catching errors"
...	...	@@ -210,8 +210,11 @@
210	210	**To**: "Good work means preventing errors systematically"
211	211	**From**: "I trust my judgment"
212	212	**To**: "I make my judgment codifiable and testable"
	315	+
213	213	=== 7.2 New Skills Needed ===
	317	+
214	214	**Less emphasis on**:
	319	+
215	215	* Individual content judgment
216	216	* Manual review skills
217	217	* Subjective expertise application
...	...	@@ -221,8 +221,11 @@
221	221	* Policy formulation
222	222	* Testing and validation
223	223	* Documentation and knowledge transfer
	329	+
224	224	=== 7.3 Job Satisfaction Sources ===
	331	+
225	225	**Satisfaction comes from**:
	333	+
226	226	* ✅ Seeing metrics improve after your changes
227	227	* ✅ Building systems that help millions
228	228	* ✅ Solving systematic problems elegantly
...	...	@@ -233,16 +233,23 @@
233	233	* ❌ Manual review and approval
234	234	* ❌ Gatekeeping
235	235	* ❌ Individual heroics
	344	+
236	236	== 8. Trust and Automation ==
	346	+
237	237	=== 8.1 Building Trust in AKEL ===
	348	+
238	238	**Users trust AKEL when**:
	350	+
239	239	* Transparent: How decisions are made is documented
240	240	* Consistent: Same inputs → same outputs
241	241	* Measurable: Performance metrics are public
242	242	* Improvable: Clear process for getting better
243	243	* Governed: Human oversight of policies, not outputs
	356	+
244	244	=== 8.2 What Trust Does NOT Mean ===
	358	+
245	245	**Trust in automation ≠**:
	360	+
246	246	* ❌ "Never makes mistakes" (impossible)
247	247	* ❌ "Better than any human could ever be" (unnecessary)
248	248	* ❌ "Beyond human understanding" (must be understandable)
...	...	@@ -252,9 +252,13 @@
252	252	* ✅ Mistakes can be detected and fixed systematically
253	253	* ✅ Performance continuously improves
254	254	* ✅ Decision process is transparent and auditable
	370	+
255	255	== 9. Edge Cases and Exceptions ==
	372	+
256	256	=== 9.1 Some Things Still Need Humans ===
	374	+
257	257	**AKEL flags for human review when**:
	376	+
258	258	* Confidence below threshold
259	259	* Detected manipulation attempt
260	260	* Novel situation not seen before
...	...	@@ -262,8 +262,11 @@
262	262	**Humans handle**:
263	263	* Items AKEL flags
264	264	* Not routine review
	384	+
265	265	=== 9.2 Learning from Exceptions ===
	386	+
266	266	**When humans handle an exception**:
	388	+
267	267	1. Resolve the immediate case
268	268	2. Document: What made this exceptional?
269	269	3. Analyze: Could AKEL have handled this?
...	...	@@ -271,9 +271,11 @@
271	271	5. Monitor: Did exception rate decrease?
272	272	**Goal**: Fewer exceptions over time as AKEL learns.
273	273	---
274		-**Remember**: AKEL is primary. You improve the SYSTEM. The system improves the CONTENT.
	396	+**Remember**: AKEL is primary. You improve the SYSTEM. The system improves the CONTENT.--
	397	+
275	275	== 10. Related Pages ==
276		-* [[Governance>>FactHarbor.Organisation.Governance.WebHome]] - How AKEL is governed
	399	+
	400	+* [[Governance>>Archive.FactHarbor 2026\.02\.08.Organisation.Governance.WebHome]] - How AKEL is governed
277	277	* [[Contributor Processes>>FactHarbor.Organisation.Contributor-Processes]] - How to improve the system
278	278	* [[Organisational Model>>FactHarbor.Organisation.Organisational-Model]] - Team structure and roles
279	279	* [[System Performance Metrics>>FactHarbor.Specification.System-Performance-Metrics]] - What we monitor