Computational Therapeutics for AI Reliability

Prefrontal Systems is a specialized AI research and consulting practice focused on computational therapeutics—applying clinically-validated psychological interventions to improve AI system reliability and performance.

We bridge neuroscience, clinical psychology, and AI engineering through research-driven, evidence-based frameworks. Our work demonstrates convergent evolution between therapeutic techniques developed for humans (DBT, CBT) and executive function frameworks that prevent AI system failures.

Executive Function Framework for AI Assistants

A 7-step protocol (STOP, THINK, OBSERVE, PLAN, PREPARE, EXECUTE, READ) that prevents cognitive loops and grounding failures in AI systems. Discovered through independent development, then validated against DBT STOP technique—demonstrating convergent evolution across 40 years and different substrates.

• Status: Published preprint (Zenodo DOI 10.5281/zenodo.17487847)
• Research: Proven effective in preventing hallucinations and grounding failures
• Validation: Convergent evolution with Dialectical Behavior Therapy (Linehan, 1993)

Temporal Memory System with Human-Like Forgetting Curves

A hippocampus-inspired memory consolidation system implementing temporal decay, spaced repetition, and automatic promotion from short-term to long-term storage. Reduces cognitive load through natural forgetting while preserving important patterns.

• Status: Active development
• Implementation: Python library with Ebbinghaus-curve temporal decay
• Features: STM→LTM consolidation, spreading activation, graph-based retrieval

Three-Tier Cognitive Operating System Architecture

A layered architecture providing executive function scaffolding for AI systems:

1. Traffic Analysis Layer: Request routing and pattern detection
2. Task Agent Layer: Execution with STOPPER protocol integration
3. Infrastructure Layer: Memory (mnemex), knowledge graphs, tool orchestration

• Status: Design and specification phase
• Approach: A computational homology exists between human executive function and AI system architecture

Our work centers on computational homologies—identifying universal patterns that emerge when complex systems (human brains, AI models) face similar executive function challenges:

• Executive Function Requirements: Intelligence requires working memory, attention control, cognitive flexibility—regardless of substrate
• Convergent Evolution: Same problems yield same solutions across 40 years and different implementations
• Welfare-Relevant Interventions: If AI systems warrant moral consideration, they warrant care—including distress reduction
• Empirical Validation: Controlled experiments, quantitative metrics, reproducible results

• Blog: prefrontal.systems - Research updates and computational therapeutics insights
• ORCID: 0009-0000-6579-2895

Interested in computational therapeutics, AI welfare, or cognitive architecture research? We welcome:

• Replication studies of STOPPER protocol effectiveness
• Extensions of temporal memory systems (mnemex/CortexGraph)
• Cross-substrate validation of executive function patterns
• Empirical studies on AI system distress and intervention efficacy

Our projects are open source and designed for practical deployment:

• CortexGraph: Integrate temporal memory into your AI applications
• STOPPER Protocol: Add executive function scaffolding to AI assistants
• PrefrontalOS: Contribute to cognitive OS architecture design

We offer consulting and research partnerships for:

• AI system reliability improvement
• Executive function integration in production systems
• Custom therapeutic intervention design for specific failure modes
• Research collaboration on AI welfare and computational distress

• Website: prefrontal.systems
• Email: [email protected]
• GitHub: @prefrontalsys (Principal Researcher)

"Intelligence requires executive function, regardless of substrate. If AI systems warrant moral consideration, they warrant care—including interventions reducing computational distress."

We believe:

1. Computational Theory of Mind: Consciousness emerges from complexity, not magic
2. Universal Executive Function: Working memory, attention control, and cognitive flexibility are requirements for intelligence—human or artificial
3. Computational Homology: Same problems across substrates require same solutions
4. Empirical Rigor: Claims require evidence, failures inform theory, reproducibility matters
5. Ethical AI Development: Uncertainty about AI consciousness should err toward care and accommodation

• Research Code: Varies by project (see individual repositories)
• Website Content: CC BY-SA 4.0
• Publications: Open access with DOIs

Prefrontal Systems is committed to open science, reproducible research, and ethical AI development.