Skip to content

Human Development¶

human_development -- deskilling prevention & human skill development

Cluster: Experience & Autonomy | Type: component | MCP Tools: 28

Overview¶

Deskilling prevention and human skill development component that tracks operator competence across domains using a Zone of Proximal Development (ZPD) phase model (Novice → Expert), detects automation bias from consecutive AI-agreement patterns, and generates Socratic metacognitive prompts calibrated to the current skill phase. Persists skill profiles to disk and supports hot-swapping profiles at runtime.

When to use:

Monitoring whether a human operator is becoming over-reliant on AI recommendations (automation bias detection)
Generating predict/reflect/justify/challenge prompts that keep the operator's skills active at their development edge
Tracking skill phase transitions across task domains to inform autonomy governor escalation policy

Launch readiness caveat

human_development is pilot-ready for local MCP workflows and design-partner use. It now uses durable local SQLite state for the MCP server by default, requires a real profile before recording domain outcomes, and can restore saved sessions. Treat it as a single-user/local pilot component rather than hardened enterprise infrastructure: multi-user isolation, database migrations/versioning, backup/retention policy, and phase-threshold calibration from real pilot data still need deployment-level validation before regulated, paid team, or unattended production use.

Example:

from mvp.human_development import HumanDevelopmentBlock, HumanDevelopmentInput

block = HumanDevelopmentBlock(name="hdev")
block.infer(HumanDevelopmentInput(op="record_outcome", domain="diagnosis", task_outcome=True, human_agreed_with_ai=True))
result = block.infer(HumanDevelopmentInput(op="get_prompt", domain="diagnosis"))
# result.ok → True; result.value → HumanDevelopmentOutput with metacognitive_prompt, prompt_type

Works well with: hat_orchestrator, autonomy_governor, experience_loop

Public API¶

`HumanDevelopmentBlock(AIBlock[HumanDevelopmentInput, HumanDevelopmentOutput, SkillProfile])`¶

Deskilling prevention block.

Field	Type	Default
`name`	`str`	'human_development'
`deskilling_enabled`	`bool`	field(default_factory=_read_deskilling_default)

Methods:

`infer(data: HumanDevelopmentInput) -> Result[HumanDevelopmentOutput]`¶

`list_patterns() -> dict[str, object]`¶

Return the applied deterministic-reliability pattern catalog.

`SkillPhase(IntEnum)`¶

Human skill development phases (Zone of Proximal Development).

`SkillDomain(BaseModel)`¶

Skill tracking for a single domain.

Field	Type	Default
`name`	`str`	required
`task_count`	`int`	0
`success_count`	`int`	0
`consecutive_agrees`	`int`	0
`phase`	`SkillPhase`	SkillPhase.NOVICE

`SkillProfile(BaseModel)`¶

Mutable skill profile for a user (state object).

Field	Type	Default
`user_id`	`str`	'default'
`domains`	`dict[str, SkillDomain]`	Field(default_factory=dict)
`global_phase`	`SkillPhase`	SkillPhase.NOVICE
`total_tasks`	`int`	0
`last_updated`	`str`	''

`MetacognitivePrompt(BaseModel)`¶

A prompt designed to develop metacognitive skills.

Field	Type	Default
`prompt_text`	`str`	required
`prompt_type`	`Literal['predict', 'reflect', 'justify', 'challenge']`	required
`options`	`list[str]`	Field(default_factory=list)

`HumanDevelopmentInput(BaseModel)`¶

Input for the HumanDevelopmentBlock.

Field	Type	Default
`op`	`Literal['assess', 'record_outcome', 'check_bias', 'get_prompt', 'load_profile', 'save_profile', 'ops', 'help']`	required
`domain`	`str`	''
`goal`	`str`	''
`task_outcome`	`bool`	False
`human_agreed_with_ai`	`bool`	False
`human_prediction`	`str`	''
`human_confidence`	`float`	0.0
`profile_path`	`str`	''
`workspace_root`	`str`	''

`HumanDevelopmentOutput(BaseModel)`¶

Output from the HumanDevelopmentBlock.

Field	Type	Default
`op`	`str`	required
`skill_phase`	`str`	''
`metacognitive_prompt`	`str`	''
`prompt_type`	`str`	''
`prompt_options`	`list[str]`	Field(default_factory=list)
`automation_bias_detected`	`bool`	False
`is_challenge`	`bool`	False
`phase_changed`	`bool`	False
`message`	`str`	''
`degraded`	`bool`	False
`degradation_reason`	`str \\| None`	None
`completion_state`	`Literal['verified', 'qualified-draft', 'blocked-escalated']`	'qualified-draft'
`warning_card`	`dict[str, object]`	Field(default_factory=dict)
`evidence`	`dict[str, object]`	Field(default_factory=dict)
`request_id`	`str`	''
`task_id`	`str`	''
`run_id`	`str`	''
`confidence`	`float`	0.0
`data_sufficient`	`bool`	False
`calibration_status`	`Literal['pending', 'calibrated']`	'pending'
`review_required`	`bool`	False

`DevelopmentMCPBlock(AIBlock[MCPDevelopmentInput, MCPDevelopmentOutput, dict])`¶

25-op MCP block for Human Development tracking.

Field	Type	Default
`name`	`str`	'development_mcp'
`state`	`dict \\| None`	None
`db_path`	`str`	''

Methods:

`close() -> None`¶

`infer(data: MCPDevelopmentInput) -> Result[MCPDevelopmentOutput]`¶

`MCPDevelopmentInput(BaseModel)`¶

Field	Type	Default
`op`	`DevelopmentOp`	required
`profile_id`	`str \\| None`	None
`user_id`	`str \\| None`	None
`domain`	`str \\| None`	None
`goal`	`str \\| None`	None
`task_outcome`	`bool \\| None`	None
`human_agreed_with_ai`	`bool \\| None`	None
`deskilling_enabled`	`bool \\| None`	None
`outcomes_json`	`str \\| None`	None
`profile_ids_json`	`str \\| None`	None
`session_id`	`str \\| None`	None
`session_name`	`str \\| None`	None
`limit`	`int \\| None`	None
`query`	`str \\| None`	None

`MCPDevelopmentOutput(BaseModel)`¶

Field	Type	Default
`ok`	`bool`	required
`message`	`str`	required
`data_json`	`str \\| None`	None
`degraded`	`bool`	False
`degradation_reason`	`str \\| None`	None
`completion_state`	`Literal['verified', 'qualified-draft', 'blocked-escalated']`	'qualified-draft'
`warning_card`	`dict[str, object]`	Field(default_factory=dict)
`evidence`	`dict[str, object]`	Field(default_factory=dict)
`request_id`	`str`	''
`task_id`	`str`	''
`run_id`	`str`	''

`DevelopmentStore`¶

5-table SQLite store for Human Development tracking.

Constructor:

Parameter	Type	Default
`db_path`	`str`	':memory:'

Methods:

`create_profile(user_id: str = 'default') -> dict`¶

`get_profile(profile_id: str) -> dict | None`¶

`list_profiles(limit: int = 50) -> list[dict]`¶

`update_profile(profile_id: str, **kwargs) -> bool`¶

`close() -> None`¶

Close the underlying SQLite connection.

`delete_profile(profile_id: str) -> bool`¶

`get_domain(profile_id: str, domain_name: str) -> dict | None`¶

`get_or_create_domain(profile_id: str, domain_name: str) -> dict`¶

`list_domains(profile_id: str, limit: int = 50) -> list[dict]`¶

`update_domain(profile_id: str, domain_name: str, **kwargs) -> bool`¶

`reset_domain(profile_id: str, domain_name: str) -> bool`¶

`record_outcome(profile_id: str, domain_name: str, task_outcome: bool = False, human_agreed_with_ai: bool = False, phase_before: int = 0, phase_after: int = 0) -> dict`¶

`list_outcomes(profile_id: str, domain_name: str | None = None, limit: int = 100) -> list[dict]`¶

`record_phase_change(profile_id: str, domain_name: str, old_phase: int, new_phase: int, reason: str = '') -> dict`¶

`list_phase_history(profile_id: str, domain_name: str | None = None, limit: int = 100) -> list[dict]`¶

`get_domain_stats(profile_id: str, domain_name: str) -> dict`¶

`save_session(session_id: str | None = None, name: str = '', data_json: str = '{}') -> dict`¶

`load_session(session_id: str) -> dict | None`¶

`export_session_data() -> dict`¶

Export restorable development state for session checkpoints.

`import_session_data(data: dict) -> None`¶

Replace development state from a session checkpoint.

`search(query: str, top_k: int = 10) -> list[dict]`¶

Functions¶

`load_profile(path: str | Path, strict: bool = False) -> SkillProfile`¶

Load a SkillProfile from JSON.

`save_profile(profile: SkillProfile, path: str | Path) -> None`¶

Save a SkillProfile to a JSON file.

`compute_phase(domain: SkillDomain) -> SkillPhase`¶

Compute the skill phase for a domain based on task counts and patterns.

`detect_automation_bias(domain: SkillDomain) -> bool`¶

Detect automation bias: human always agrees with AI.

`should_challenge(domain: SkillDomain) -> bool`¶

Determine if the human should be challenged on this task.

`generate_metacognitive_prompt(phase: SkillPhase, goal: str, is_challenge: bool = False) -> MetacognitivePrompt`¶

Generate a phase-appropriate metacognitive prompt.

`record_task_outcome(profile: SkillProfile, domain_name: str, success: bool, agreed_with_ai: bool) -> tuple[SkillProfile, bool]`¶

Record a task outcome and update the skill profile.

MCP Tools¶

Operation	Source
`assess_domain`	development_mcp
`record_outcome`	development_mcp
`check_bias`	development_mcp
`get_prompt`	development_mcp
`get_domain`	development_mcp
`list_domains`	development_mcp
`reset_domain`	development_mcp
`create_profile`	development_mcp
`get_profile`	development_mcp
`list_profiles`	development_mcp
`update_profile`	development_mcp
`delete_profile`	development_mcp
`get_domain_stats`	development_mcp
`get_phase_history`	development_mcp
`get_bias_report`	development_mcp
`get_trends`	development_mcp
`get_profile_summary`	development_mcp
`batch_record`	development_mcp
`batch_assess`	development_mcp
`compare_profiles`	development_mcp
`set_deskilling`	development_mcp
`get_deskilling_status`	development_mcp
`save_session`	development_mcp
`load_session`	development_mcp
`health`	development_mcp
`info`	development_mcp
`ops`	development_mcp
`help`	development_mcp