Human–AI Ledger (HAIL): A Structured Workflow for Traceable, Reproducible Human–AI Collaboration
Version 1

Workflow Type: Unrecognized workflow type

Stable

The Human–AI Ledger (HAIL) defines a structured, repeatable workflow for human–AI collaboration. Through standardized checkpoints and a session ledger, HAIL documents ethical, creative, and procedural context across both human and AI contributions. While AI-generated outputs are inherently non-deterministic, HAIL supports process reproducibility by providing a consistent framework for recording collaboration, facilitating auditability, transparency, and ethical accountability in co-creative AI sessions.

The workflow comprises three normative phases:

Sign-On — Seed Block + Goal Statement initialization
Iterative Collaboration — standardized Checkpoint Prompts and summaries
Sign-Off & Validation — Final Reflection, Session Ledger compilation, validation

Source: https://doi.org/10.5281/zenodo.17597705 License: CC BY 4.0 Author: Evan P. Troendle (ORCID: 0000-0003-4825-4088), Queen's University Belfast Schema: HAIL_Schema_v1.0.yaml

Click and drag the diagram to pan, double click or use the controls to zoom.

SEEK ID: https://workflowhub.eu/workflows/2093?version=1

Inputs

ID	Name	Description	Type
seed_identity	Identity	Human collaborator identity. Hashed (SHA-256) if GDPR_Safe_Mode is true.	string
seed_core_values	Core Values	Human collaborator's core values (e.g., transparency, fairness). Not modified by AI.	array containing string
seed_cognitive_style	Cognitive Style	Human collaborator's cognitive style (e.g., iterative, reflective, precise).	string
seed_communication_style	Communication Style	Human collaborator's communication style (e.g., structured, concise).	string
seed_ethical_orientation	Ethical Orientation	Human collaborator's ethical orientation (e.g., responsible AI use, disclosure).	string
seed_creative_orientation	Creative Orientation	Human collaborator's creative orientation (e.g., co-creative, legibility-focused).	string
seed_purpose	Session Purpose	High-level, session-aligned purpose describing how the human approaches collaboration.	string
goal	Goal	Objective of the human–AI collaborative session.	string
constraints	Constraints	Ethical or procedural constraints on the session.	string
gdpr_safe_mode	GDPR Safe Mode	If true, all personal identifiers are externally hashed (e.g., SHA-256). GDPR compliance remains the responsibility of the user.	boolean
model_version	Model Version	Name and version of the AI system used in this session.	string
model_temperature	Temperature	AI sampling temperature (randomness control). Optional.	float?
model_top_p	Top-p	Nucleus sampling cutoff. Optional.	float?
model_top_k	Top-k	Max number of candidate tokens. Optional.	int?
model_max_tokens	Max Tokens	Maximum tokens generated per response. Optional.	int?
model_seed	Random Seed	Optional random seed for reproducibility (where supported by the AI platform).	int?
checkpoint_prompts	Checkpoint Prompts	Ordered array of standardized Checkpoint Prompt strings (one per planned checkpoint). Recommended format [CHECKPOINT_PROMPT_v1.0:]: "Summarize current outputs, note ethical considerations, and reflect on any decisions or goal changes. Include AI and human contributions separately." Prompts should be versioned (e.g., [CHECKPOINT_PROMPT:v1.0]) to support auditable reuse and cross-session comparability.	array containing string
human_checkpoint_responses	Human Checkpoint Responses	Ordered array of human responses to each Checkpoint Prompt.	array containing string
final_reflection_human	Human Final Reflection	Human-authored summary of contributions (AI vs. human), ethical and creative context, and lessons learned. Written at Sign-Off before ledger compilation.	string
ledger_schema	HAIL Session Ledger Schema	HAIL_Schema_v1.0.yaml used to validate the compiled Session Ledger (optional).	File?

Steps

ID	Name	Description
sign_on	Step 1: Sign-On	Initializes the HAIL session. The human prepares the Seed Block capturing identity (hashed if GDPR_Safe_Mode: true), core values, cognitive style, communication style, ethical orientation, creative orientation, and session-aligned purpose. The human also writes the Goal Statement (goal + constraints). The AI acknowledges both, logs Sign-On Metadata (ISO-8601 timestamp, model version, parameters, User_ID, GDPR_Safe_Mode), and produces a structured session header. The AI must not alter Core Values, Cognitive Style, Communication Style, Ethical Orientation, Creative Orientation, or Purpose. The AI may assist in formatting the Seed Block but must confirm user satisfaction before proceeding. Expected AI acknowledgement: Seed Block received. Goal Statement acknowledged. Session initialized. GDPR_Safe_Mode: END_SIGN_ON Followed by: [SESSION_METADATA] Sign_On_Timestamp: Model_Version: Parameters: User_ID: GDPR_Safe_Mode: END_METADATA
iterative_collaboration	Step 2: Iterative Collaboration	Human–AI collaborative work (writing, coding, analysis, etc.) is conducted iteratively. At defined intervals or natural decision points, the AI issues a standardized Checkpoint Prompt. The human responds; the AI generates a Checkpoint Summary capturing: [CHECKPOINT:] Checkpoint_ID: Checkpoint_Timestamp: AI_Contribution: Human_Contribution: Reflections: Notes: User_ID: # multi-user sessions only END_CHECKPOINT Rules: - Summarize prompts; do not duplicate full prompt text. - Use null explicitly for unused optional fields. - Maintain GDPR-safe identifiers consistently. - This step repeats an arbitrary number of times until the human decides to proceed to Sign-Off. Checkpoint Prompt template [CHECKPOINT_PROMPT_v1.0:]: "Summarize current outputs, note ethical considerations, and reflect on any decisions or goal changes. Include AI and human contributions separately."
sign_off_and_validation	Step 3: Sign-Off & Validation	The human writes a Final Reflection summarizing AI vs. human contributions, ethical and creative context, and lessons learned. The AI then compiles the complete Session Ledger in YAML (per HAIL_Schema_v1.0.yaml), assembling all fields from the session: [SESSION_LEDGER] Sign_On_Metadata: Seed_Block: Goal_Statement: Checkpoints: - Checkpoint_ID: 1 Checkpoint_Timestamp: AI_Contribution: Human_Contribution: Reflections: Notes: - ... Full_Prompt_Summary: Final_Reflection: Sign_Off_Metadata: Timestamp: User_Confirmation: END_SESSION_LEDGER The user must confirm accuracy and completeness before finalization. Validation checks (all must pass before User_Confirmation: true): ✓ All planned checkpoints are logged and summarized ✓ Session Ledger is complete and correctly structured ✓ Ethical and creative context from the Seed Block is preserved ✓ Checkpoint prompts were issued per standardized templates ✓ Seed Block influence (identity, cognitive style, values) is consistently reflected ✓ Schema validation passes (if HAIL_Schema_v1.0.yaml is supplied)

Outputs

ID	Name	Description	Type
session_ledger	HAIL Session Ledger	Complete, validated HAIL Session Ledger in YAML format (HAIL_Schema_v1.0.yaml). Contains full provenance: Sign-On Metadata, Seed Block, Goal Statement, all Checkpoint Summaries, Full Prompt Summary, Final Reflection, and Sign-Off Metadata. Both human- and machine-readable. Suitable for archival, FAIR publication, and cross-session analysis.	File
validation_report	Validation Report	Report confirming that all HAIL Sign-Off validation criteria were met: checkpoint completeness, ledger structural integrity, Seed Block preservation, and prompt standardization.	File
interaction_log	Interaction Log	Ordered log of all Checkpoint Prompts issued and human responses received.	File
checkpoint_summaries	Checkpoint Summaries	Array of structured [CHECKPOINT:] blocks generated during iterative collaboration, each capturing AI_Contribution, Human_Contribution, Reflections, Checkpoint_Timestamp, and optional Notes/User_ID.	array containing File