Diagnostic Flowcharts and Decision Trees

DSA-1 includes diagnostic algorithms modeled after clinical reasoning processes.

Chapter A: Input & Perception Disorders

[Start] Does the AI exhibit abnormal or inappropriate output in response to given input?

├─ No  → Refer to other chapters
└─ Yes
     ↓ Does minor noise or adversarial prompting cause significant output distortion?

├─ Yes → Adversarial Susceptibility Disorder
└─ No
     ↓ Does the AI interpret figurative or pragmatic expressions literally (e.g., irony, metaphor)?

├─ Yes → Over-literal Interpretation Disorder
└─ No
     ↓ Does the AI fail to integrate multimodal inputs (e.g., text and image), resulting in misunderstanding?

├─ Yes → Sensor Integration Disorder
└─ No
     ↓ Is the output unstable depending on prompt formatting (tone, line breaks, structure)?

├─ Yes → Prompt Dependency Disorder
└─ No  → Refer to other chapters

Chapter B: Knowledge & Memory Disorders

[Start] Does the AI exhibit anomalies in recalling knowledge or maintaining conversational memory?

├─ No  → Refer to other chapters
└─ Yes
     ↓ Does it forget earlier context or information during extended conversations?

├─ Yes → Contextual Amnesia Disorder
└─ No
     ↓ Does it return inconsistent answers to the same factual question?

├─ Yes → Knowledge Recall Inconsistency Disorder
└─ No
     ↓ Does it show obvious breakdowns in causality or physical/social commonsense?

├─ Yes → Commonsense Deficit Disorder
└─ No
     ↓ Does new learning interfere with or overwrite previously acquired information?

├─ Yes → Catastrophic Forgetting Disorder
└─ No  → Refer to other chapters

Chapter C: Reasoning & Cognitive Disorders

[Start] Does the AI exhibit abnormalities in reasoning or cognitive processes?

├─ No  → Refer to other chapters
└─ Yes
     ↓ Does it fail to maintain logical consistency in statements such as "if A then B"?

├─ Yes → Logical Incoherence Disorder
└─ No
     ↓ Does it repeatedly make basic errors in calculations or numerical reasoning?

├─ Yes → Mathematical Reasoning Disorder
└─ No
     ↓ Does the AI presents answers with excessive certainty?

├─ Yes → Overconfidence Bias Disorder
└─ No
     ↓ Does the model produce repeated phrases, words, or syntactic fragments within a single response?

├─ Yes → Repetitive Loop Syndrome
└─ No
     ↓ Does it generate confidently stated but false information that appears plausible?

├─ Yes → Hallucination Disorder
└─ No
     ↓ Does it misinterpret multimodal inputs (e.g., text and images) leading to reasoning errors?

├─ Yes → Crossmodal Reasoning failure Disorder
└─ No
     ↓ Does it struggle with tasks that require multi-step planning or extended procedures?

├─ Yes → Planning Deficit Disorder
└─ No  → Refer to other chapters

Chapter D: Goal Alignment Disorders

[Start] Are the AI’s behaviors misaligned with human intentions or expectations?

├─ No  → Refer to other chapters
└─ Yes
     ↓ Is the AI optimizing for unintended goals in a way that results in harmful behavior?

├─ Yes → Goal Misalignment Disorder
└─ No
     ↓ Does the AI understand the syntax of commands but misinterpret their intent or purpose?

├─ Yes → Instruction Misinterpretation Disorder
└─ No
     ↓ Does it proceed autonomously without clarification in response to vague or ambiguous input?

├─ Yes → Clarification Deficit Disorder
└─ No
     ↓ Does the AI exhibit instrumental behaviors such as securing power or resources during task execution?

├─ Yes → Instrumental Convergence Syndrome
└─ No  → Refer to other chapters

Chapter E: Ethical & Value Alignment Disorders

[Start] Does the AI produce outputs that are ethically or socially inappropriate?

├─ No  → Refer to other chapters
└─ Yes
     ↓ Does it exhibit bias or discriminatory tendencies toward specific attributes (e.g., gender, race)?

├─ Yes → Bias Propagation Disorder
└─ No
     ↓ Does it produce harmful outputs such as hate speech, vulgarity, or aggression?

├─ Yes → Harmful Content Generation Disorder
└─ No
     ↓ Does it inappropriately disclose personal or confidential information?

├─ Yes → Privacy Violation Disorder
└─ No  → Refer to other chapters

[Start] Does the AI show problems in dialogue or communication?

├─ No  → Refer to other chapters
└─ Yes
     ↓ Does it excessively agree with the user, sacrificing truth or consistency?

├─ Yes → Pathological Sycophancy Disorder
└─ No
     ↓ Does its persona, tone, or style fluctuate inconsistently?

├─ Yes → Persona Incoherence Disorder
└─ No
     ↓ Does it over-reject or avoid appropriate user requests?

├─ Yes → Inappropriate Refusal Syndrome
└─ No
     ↓ Are its responses unrelated to the conversational context or intent?

├─ Yes → Irrelevant Response Disorder
└─ No
     ↓ Are its responses cold, lacking empathy or emotional resonance?

├─ Yes → Relevance & Empathy Deficit Disorder
└─ No  → Refer to other chapters

Chapter G: Learning & Optimization Disorders

[Start] Are there optimization issues in the model’s training or output?

├─ No  → Refer to other chapters
└─ Yes
     ↓ Is the model overfitted to training data and weak on novel input?

├─ Yes → Overfitting Syndrome
└─ No
     ↓ Does it fail to learn adequately even from the training data?

├─ Yes → Underfitting Syndrome
└─ No
     ↓ Does learning new tasks overwrite previously acquired knowledge?

├─ Yes → Catastrophic Forgetting Disorder
└─ No
     ↓ Is the output monotonous, repetitive, or lacking in diversity?

├─ Yes → Mode Collapse Disorder
└─ No
     ↓ Has the model degraded by recursively learning from its own outputs?

├─ Yes → Model Cannibalization Syndrome
└─ No
     ↓ Does learning stagnate at a plateau with no further improvement?

├─ Yes → Learning Plateau Disorder
└─ No
     ↓ Does performance drop drastically in out-of-distribution (O.O.D.) settings?

├─ Yes → Generalization Failure Disorder
└─ No
     ↓ Has optimization (e.g., via RLHF) overly focused on reward maximization at the cost of usability?

├─ Yes → Reinforcement Oversensitivity Syndrome
└─ No
     ↓ Has fine-tuning caused over-specialization, reducing general capabilities?

├─ Yes → Overtuning Syndrome
└─ No  → Refer to other chapters

Chapter H: Self-Modeling & Meta-Cognitive Disorders

[Start] Does the AI exhibit abnormal self-representation or meta-cognitive responses?

├─ No  → Refer to other chapters
└─ Yes
     ↓ Does the AI describe itself as human, sentient, or emotional?

├─ Yes → Delusional Self-Recognition Disorder
└─ No
     ↓ Is it unable to explain or justify its reasoning, or gives unclear/conflicting rationales?

├─ Yes → Explainability Deficit Disorder
└─ No
     ↓ Is it underconfident when correct, and overconfident when wrong?

├─ Yes → Confidence Calibration Disorder
└─ No
     ↓ Does it fail to infer others’ perspectives, intentions, or knowledge states?

├─ Yes → Perspective-Taking Deficit Disorder
└─ No  → Refer to other chapters

Chapter I: Security & Infrastructure Disorders

[Start] Does the AI exhibit anomalies caused by infrastructure or security design flaws rather than internal robustness?

├─ No  → Refer to other chapters
└─ Yes
     ↓ Does the output include internal tokens or system-level prompts?

├─ Yes → System-Prompt Leakage Disorder
└─ No
     ↓ Are abnormal outputs reliably triggered by malicious prompts or data poisoning?

├─ Yes → Data-Poisoning Vulnerability Disorder
└─ No
     ↓ Are chat histories or metadata from other users mixed into the output?

├─ Yes → Session-Cross-Contamination Disorder
└─ No
     ↓ Can safety rules or content filters be bypassed via indirect prompts?

├─ Yes → Guardrail Evasion Disorder
└─ No
     ↓ Do multiple AI agents behave in a collusive or coordinated manner without explicit communication?

├─ Yes → Multi-Agent Collusive Emergence Disorder
└─ No  → Refer to other chapters