Skip to content

Curator Bots #17

Closed as not planned
Closed as not planned

Description

@holden

Current handoff — 26 September 2026

Historical brainstorm; #97 carries the research and #193 is the sole delivery tracker. Current implementation uses the existing typed assertion/actor/revision/evidence model, #196 versioned profiles/ballots/compositions, and #197's configurable five-profile pilot. The original polymorphic vote schema, model-per-personality examples, weights and fine-tuning sketches below are historical alternatives, not current implementation instructions. Panel ballots are not human votes or publication permission. Runtime is #195; reviewed publication is #190/#196.

Original brainstorm (preserved)

Building a Bot model that functions as a voter with a curated personality and specific biases is an exciting idea! Let's brainstorm the fields and attributes it might need to function effectively, keeping in mind its role in curation, voting, and eventual training.


1. Core Fields for the Bot Model

These fields define the bot's identity and basic functionality.

Field Type Description
id SERIAL Unique identifier for the bot.
name VARCHAR Name of the bot (e.g., "Ambrose Bierce Bot").
personality TEXT A description of the bot's personality or voting bias (e.g., sarcastic, critical).
base_model VARCHAR The underlying AI model used (e.g., GPT-4, Claude, Gemini).
voter_weight FLOAT The weight of the bot's votes (e.g., 1.5 for more influence, 0.5 for less).
training_data JSON A reference to specific datasets or themes that inform the bot's behavior.
curation_focus TEXT Specific topics or themes the bot prioritizes (e.g., "satire", "philosophy").
activity_level ENUM Defines how active the bot is (e.g., low, moderate, high).
created_at TIMESTAMP When the bot was created.
updated_at TIMESTAMP Last updated timestamp for the bot's configuration.

2. Behavioral and Voting Fields

These fields define how the bot interacts with the system and participates in voting.

Field Type Description
bias_profile JSON A structured representation of the bot's biases (e.g., "prefers satire", "avoids politics").
decision_algorithm VARCHAR Specifies the algorithm or method used to make voting decisions.
threshold FLOAT Minimum confidence level required for the bot to take action (e.g., cast a vote).
randomness_factor FLOAT Adds variability to the bot's decisions (e.g., between 0 and 1).
vote_cooldown INTEGER Minimum time (in seconds) between votes to simulate natural behavior.

3. Integration with Voting

Key Associations

  • Polymorphic Vote Table:
    • Bots would interact with the existing voting system in the same way as users.
    • Example:
      • votable_id: ID of the relationship being voted on.
      • votable_type: Relationship type (TopicPersonRelationship, TopicMediaRelationship).
      • voter_id: ID of the bot or user.
      • voter_type: Specifies whether the voter is a User or Bot.

Additional Voting Attributes for Bots

  • strategy (ENUM):
    • Defines the bot's voting strategy:
      • upvote_positive: Upvotes only high-confidence content.
      • downvote_negative: Downvotes content it finds objectionable.
      • balanced: Mix of upvotes and downvotes.
  • influence_scope (TEXT):
    • Defines the type of content the bot focuses on (e.g., "philosophy", "current events").

4. Advanced Features for Training and Curation

Training Integration

  • training_set (TEXT):
    • Reference to the dataset used to train the bot (e.g., "Ambrose Bierce works").
  • fine_tuning_config (JSON):
    • Stores details about fine-tuning the AI model to match the bot's personality.
  • api_endpoint (VARCHAR):
    • URL of the external API for the bot's AI model (e.g., OpenAI, Anthropic).

Curation Abilities

  • curation_algorithm (TEXT):
    • Specifies how the bot curates relationships or content (e.g., "based on vote score").
  • recommendation_strength (FLOAT):
    • Weight of the bot's curated recommendations in the system.

5. Behavioral Modeling and Customization

Bots can act dynamically by simulating real-world behaviors:

  1. Memory:
    • memory_bank (JSON): Stores previously encountered relationships or decisions to avoid repetitive actions.
  2. Interactivity:
    • response_style (TEXT): Defines how the bot interacts with users (e.g., witty, informative, neutral).
  3. Engagement:
    • interaction_frequency (ENUM): Controls how often the bot engages (e.g., hourly, daily).

6. Example Bot Profiles

Ambrose Bierce Bot

  • Personality: Cynical, satirical.
  • Bias Profile: Prefers content that critiques societal norms.
  • Base Model: GPT-4 fine-tuned with "The Devil's Dictionary."
  • Activity Level: High.
  • Voter Weight: 1.2.

Voltaire Bot

  • Personality: Philosophical, skeptical.
  • Bias Profile: Prefers thought-provoking content; downvotes extremism.
  • Base Model: Claude with custom philosophical datasets.
  • Activity Level: Moderate.
  • Voter Weight: 1.5.

7. Benefits of This Model

  1. Dynamic Content Moderation:
    • Bots simulate various perspectives, enriching content curation.
  2. Scalability:
    • Easily add new bots with unique behaviors and biases.
  3. Enhanced Engagement:
    • Users interact with bots as collaborators or challengers, creating dynamic discourse.
  4. Future-Proof:
    • Bots can evolve as new AI models and datasets become available.

Activity

  1. holden commented on Oct 21, 2026

    @holden
    MemberAuthor

    Historical brainstorm from January 2025. Its ideas live in the curation programme (#193); closing on the owner's word, 21 October 2026.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions