Skip to content

Commit 6744880

Browse files
committed
Docs: Implement Tone Damping, Cultural Calibration, and The Accountability Handbook
1 parent b50d601 commit 6744880

3 files changed

Lines changed: 44 additions & 2 deletions

File tree

docs/guides/managers.md

Lines changed: 33 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,33 @@
1+
# The Accountability Handbook: A Guide for Managers 👔
2+
3+
CommitGuard AI is a tool of empowerment, not surveillance. For a Human-in-the-Loop (HITL) system to work, managers must know how to translate AI flags into professional leadership.
4+
5+
---
6+
7+
## 1. Interpreting the Flags
8+
When CommitGuard flags an **"Ambiguity"** or a **"Truth Gap,"** it is a prompt for a conversation, not a citation for failure.
9+
10+
* **Flag: "Ambiguous Commitment"**
11+
* *AI Logic*: The user's words were too vague to extract a deadline.
12+
* *Manager Action*: Don't scold. Instead, reply: *"Hey [Name], I saw your update. To help me clear any blockers, could you give me a rough target for when you'll have this in review?"*
13+
* *Goal*: Precision, not pressure.
14+
15+
* **Flag: "Truth Gap Detected"**
16+
* *AI Logic*: Verbal claims don't match Git evidence.
17+
* *Manager Action*: Assume there is a technical blocker. Ask: *"I noticed the repo hasn't updated since our chat. Are there any local environment issues or merge conflicts I should jump in on?"*
18+
* *Goal*: Removing blockers, not investigating honesty.
19+
20+
---
21+
22+
## 2. Managing the "Human-in-the-Loop" (HITL)
23+
High-stakes interventions should never be 100% autonomous.
24+
25+
1. **Review the Confidence Score**: If the AI has a `confidence_score` of 0.8 but flags a high risk, verify if the user's recent Slack messages were sarcastic or high-context.
26+
2. **Calibrate for Morale**: If the team has just finished a "Death March" (long sprint), manually increase the **Supportive Threshold** to allow for natural recovery.
27+
3. **Cross-Cultural Norms**: Ensure your specific team's "Directness Level" is reflected in the agent's tone settings.
28+
29+
---
30+
31+
## 3. Best Practices for Professionalism
32+
* **Public vs. Private**: CommitGuard monitors public channels, but follow-up interventions for **high-risk signals** should often be moved to a 1-on-1 private DM thread once the manager is alerted.
33+
* **Celebrate the Reliability**: When someone consistently hits 100% on their Reliability Score, use the report as a basis for formal recognition. Accountability is as much about **Praise** as it is about Progress.

docs/overviews/ethics.md

Lines changed: 9 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -9,9 +9,16 @@ One of our most discussed features is the **"Confrontational"** tone. Here is ho
99

1010
* **The Empathy Buffer**: The system is hardcoded to prioritize `SUPPORTIVE` tones if anything resembling burnout or personal distress is detected.
1111
* **The Burnout Safety Valve**: If the `ExcuseDetector` identifies signs of fatigue, the system **blocks** confrontational escalation and triggers a "Burnout Alert" for the manager instead.
12-
* **Tone as Calibration, Not Punishment**: The objective isn't to "scold" but to match the professional urgency required by the deadline.
12+
* **Tone Drift & Cooling-off**: To prevent morale fatigue, the system implements **tone-damping**. If a user has received multiple "Firm" follow-ups, the system enters a 48-hour "Cooling-off" state where only supportive or neutral tones are permitted to avoid wearing down the individual.
13+
14+
## 2. Cultural & Contextual Sensitivity
15+
"Deflection" is relative. What is seen as blunt in one culture is polite in another:
16+
17+
* **Sensitivity Calibration**: CommitGuard supports **Cultural Tone Profiles**. Managers can calibrate the "Pressure Sensitivity" of the agents to match their specific team norms (e.g., High-Directness vs. High-Context locales).
18+
* **Domain-Specific Jargon**: The NLP models are refined to recognize that certain industry vernacular (e.g., *"I'm swamped"*) may be a routine status update rather than an excuse in specific high-velocity teams.
19+
20+
## 3. Privacy & Data Integrity
1321

14-
## 2. Privacy & Data Integrity
1522
Monitoring at the granularity of Slack threads and Git commits requires a strict privacy stance:
1623

1724
* **Scoped Monitoring**: CommitGuard is designed to monitor designated `#project` channels, not private DMs or unrelated chatter.

mkdocs.yml

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -60,6 +60,8 @@ nav:
6060
- Quick Start: guides/quickstart.md
6161
- Slack Integration: guides/slack.md
6262
- GitOps Accountability: guides/gitops.md
63+
- The Accountability Handbook: guides/managers.md
64+
6365

6466
- API Reference:
6567
- Endpoint Specs: reference/api.md

0 commit comments

Comments
 (0)