Forward Mode

Default temporal mode with strict causality and knowledge provenance

What is Forward Mode?

Forward mode is the default temporal causality in Timepoint Pro. Events flow from origin to endpoint in strict chronological order, with rigorous enforcement of causal dependencies and knowledge provenance.

Think of Forward mode as Pearl causality: events occur because prior events caused them, and entities can only know what they’ve been explicitly exposed to.

Core principle: No anachronisms. An entity cannot reference knowledge they haven’t yet acquired through an exposure event.

When to Use Forward Mode

Use Forward mode when you need:

• Realistic causality - Events must follow plausible cause-and-effect chains
• Knowledge tracking - Who knows what, when, and from whom matters
• Legal/compliance scenarios - Discovery timelines, information barriers, audit trails
• Training data - Generate datasets where temporal logic must be sound
• Default behavior - When no special temporal reasoning is needed

Best for

• Board meetings
• Discovery timelines
• Regulatory audits
• Contract negotiations
• Multi-party investigations

Not ideal for

• Root cause analysis (use Portal)
• Counterfactual “what-ifs” (use Branching)
• Prophecy/fate narratives (use Cyclical)
• Dramatic storytelling (use Directorial)

How Forward Mode Works

1. Define Origin

   Specify the starting state with origin_year and initial conditions.

   {
     "temporal": {
       "mode": "forward",
       "origin_year": 2025
     }
   }
   
2. Generate Consequents

   At each step, the system generates plausible next states based on:

   • Current entity states and capabilities
   • Available knowledge
   • Causal necessity (what must follow given prior events)
   • Resource constraints
3. Enforce Provenance

   Knowledge flows are tracked via exposure events (M3):

   • entity_a learns fact_x from entity_b at timepoint_5
   • entity_a cannot reference fact_x before timepoint_5
   • Graph edges capture who-knows-what chains
4. Validate Causality

   Each forward step is validated for:

   • Temporal consistency (no future knowledge)
   • Entity capability (can they perform this action?)
   • Resource availability (are required inputs present?)

Configuration

Forward mode is the default. You can configure it explicitly in your template:

{
  "temporal": {
    "mode": "forward",
    "origin_year": 2025,
    "backward_steps": 10,  // actually forward steps
    "coherence_threshold": 0.7,
    "candidate_antecedents_per_step": 3
  }
}

Configuration Parameters

Parameter	Type	Default	Description
mode	string	"forward"	Must be "forward" (or omit for default)
origin_year	int	required	Starting year for simulation
backward_steps	int	10	Number of forward timepoints to generate
coherence_threshold	float	0.7	Minimum plausibility score (0.0-1.0)
candidate_antecedents_per_step	int	3	Candidate futures per step
exploration_mode	string	"adaptive"	Strategy for path exploration

Template Examples

Example 1: Board Meeting (Standard)

From showcase/board_meeting.json:

{
  "scenario_description": "Tech startup board meeting with CEO acquisition proposal. 4 entities (CEO, lead investor, board member, advisor) discuss strategic options. Dialog synthesizes with relationship tracking.",
  "temporal": {
    "mode": "forward"  // implicit - this is default
  },
  "metadata": {
    "mechanisms_featured": [
      "M1_heterogeneous_fidelity",
      "M7_causal_chains",
      "M11_dialog_synthesis",
      "M13_relationship_tracking"
    ]
  }
}

Cost: $0.05 | Duration: ~2min | Entities: 4 | Timepoints: 3-5

Example 2: Litigation Discovery (Forward with M19)

From persona/agent3_litigation_discovery.json:

{
  "scenario_description": "FORWARD mode litigation with enforced information asymmetry through discovery timeline. Plaintiff counsel gains knowledge sequentially through document production, depositions, and expert reports.",
  "temporal": {
    "mode": "forward",
    "origin_year": 2023
  },
  "metadata": {
    "mechanisms_featured": [
      "M3_exposure_events",
      "M7_causal_chains",
      "M11_dialog_synthesis",
      "M13_relationship_tracking",
      "M19_information_asymmetry"  // Core mechanic
    ]
  }
}

Cost: $0.06 | Duration: ~2min | Use case: Discovery timeline simulation

Knowledge provenance in action: Plaintiff’s attorney cannot reference the smoking-gun email until it’s produced in discovery at timepoint 4, even though defense counsel knew about it from timepoint 1.

Code: run() Method Signature

Forward mode uses the default workflow in workflows/temporal_agent.py. There’s no dedicated ForwardStrategy class because Forward is the baseline behavior.

The main orchestrator is:

# workflows/temporal_agent.py
class TemporalAgent:
    def run_simulation(
        self,
        config: TemporalConfig,
        query_sequence: list[Query]
    ) -> SimulationResult:
        """
        Default forward simulation with causal validation.
        
        Process:
        1. Initialize entities at origin_year with initial_knowledge
        2. For each forward step:
           - Generate candidate next states
           - Validate causal consistency
           - Enforce knowledge provenance (M3)
           - Update entity states
        3. Return timeline with exposure event graph
        """

Best Practices

1. Seed Initial Knowledge Carefully

What entities know at t=0 determines what’s possible downstream.

{
  "entity_roster": {
    "ceo_sarah": {
      "initial_knowledge": [
        "company_financials_q4_2024",
        "acquisition_offer_30m",
        "board_member_names"
      ]
    }
  }
}

Why: Entities can only reference facts they’ve been exposed to.

2. Use Exposure Events for Critical Knowledge

For legally/causally sensitive information, create explicit exposure events:

{
  "timepoints": [
    {
      "event_description": "Document production: Email thread reveals price-fixing discussion",
      "exposure_events": [
        {
          "source": "defense_counsel",
          "target": "plaintiff_counsel",
          "knowledge_transferred": "smoking_gun_email_thread"
        }
      ]
    }
  ]
}

3. Set Coherence Threshold Based on Scenario

• High-stakes legal/compliance: coherence_threshold: 0.8 (strict)
• Exploratory ideation: coherence_threshold: 0.6 (permissive)
• Training data generation: coherence_threshold: 0.7 (balanced)

4. Leverage Relationship Tracking (M13)

Forward mode works beautifully with M13_relationship_tracking:

• Trust evolves based on prior interactions
• Tension accumulates from conflicts
• Dialog tone reflects relationship state

Enable in template:

{
  "metadata": {
    "mechanisms_featured": ["M13_relationship_tracking"]
  }
}

Cost Estimates

Quick

$0.02 -$0.05

2-3 entities

1-2 timepoints

Under 1 min

Standard

$0.05 -$0.20

4-6 entities

3-5 timepoints

1-5 min

Comprehensive

$0.20 -$1.00

7-10 entities

5-10 timepoints

5-15 min

Related Mechanisms

Forward mode commonly pairs with:

• M3 (Exposure Events) - Knowledge provenance tracking
• M7 (Causal Chains) - Event ordering validation
• M11 (Dialog Synthesis) - Multi-turn conversation generation
• M13 (Relationship Evolution) - Trust/tension dynamics
• M19 (Information Asymmetry) - Enforced knowledge barriers

Next: Explore Other Modes

• Portal Mode - Backward reasoning from outcomes
• Branching Mode - Counterfactual timelines
• Cyclical Mode - Repeating patterns and prophecy
• Directorial Mode - Narrative-driven arcs