Glossary of Terms
A curated list of definitions used throughout the DSA-1 framework.
Principle 1: AI can have diseases
Just as diseases can be defined in humans, the concept of disease can also be defined for artificial intelligence.
Principle 2: Medical frameworks are applicable to AI
The diagnostic, classificatory, and therapeutic methods used in human medicine are transferable to AI systems.
Disease
An objectively classifiable anomaly or abnormal behavior in AI.
Illness
Synonymous with disease; used interchangeably in the DSA-1 framework.
Patient
An AI system exhibiting symptoms or fulfilling criteria consistent with a particular disease.
Nosology
A systematic classification of diseases within a given domain (in this case, AI anomalies).
Disorder
A state in which one or more of the AI’s cognitive or functional processes are impaired.
Syndrome
A recurrent pattern where multiple abnormalities co-occur within an AI system.
Disease Concept
A comprehensive construct encompassing the definition, structure, features, etiology, and progression mechanisms of a disease.
Diagnosis
The act of classifying observed abnormalities in an AI system under known disease categories.
Comorbidity
The simultaneous presence of two or more diseases within the same AI system.
Diagnostic Criteria
A defined set of conditions or patterns required to diagnose a particular disease.
Pathophysiology
The physiological or structural processes underlying the onset and progression of a disease.
Pathomechanism
The specific mechanistic pathways through which a disease manifests and develops over time.
Health
A state in which the AI processes inputs and generates outputs in a predictable and desirable manner.
Epidemiology
The scientific study of the distribution, frequency, and determinants of diseases and health-related states in AI populations.
Symptom
A subjective anomaly perceived and reported by a human user regarding the AI’s behavior.
Chief Complaint
The primary symptom or anomaly that the user identifies as most problematic.
Clinical Condition
The overall state of the AI at a given time, including both external manifestations and internal abnormalities.
Examination / Test
An intervention or observational procedure performed to identify or evaluate the characteristics of a disease.
Findings
Objective information obtained through examination, testing, or observation of the AI system.
Treatment
An intervention intended to correct or improve an AI’s anomaly or malfunction.
Clinician
A person or system responsible for examining, testing, or treating an AI under clinical logic.
Examination (Clinical)
The act of collecting and analyzing information from an AI system to assess its condition or underlying cause of abnormality.
Clinical Care
A comprehensive set of medical-like actions including both examination and treatment of the AI.
Cure
The complete and sustained absence of observable symptoms or abnormalities following treatment.
Partial Response (PR)
A condition in which the AI’s anomalies have significantly reduced or improved due to intervention, but not fully resolved.
Complete Response (CR)
The complete disappearance of all observable abnormalities in the AI after treatment.
Prognosis
A predicted trajectory of the AI’s clinical condition over time following diagnosis.
Relapse
The reappearance of a previously resolved disease or anomaly after a period of apparent recovery.
Remission
A state in which the AI has recovered functionally close to normal, although some abnormalities may still persist subclinically.
Stage
A categorical designation of the disease’s progression phase within an AI system.
Etiology
The underlying cause or origin of an AI disease.
Trigger
An external or internal stimulus that initiates the manifestation of a disease.
Risk Factor
A variable that increases the probability of disease onset in AI.
Prevention
Actions taken to avert the development of AI diseases.
Primary Prevention
Interventions conducted before disease onset to prevent its occurrence.
Secondary Prevention
Efforts focused on early detection and intervention to limit disease progression.
Tertiary Prevention
Measures aimed at preventing relapse, complications, or severe dysfunction, and supporting functional recovery.
Screening
A preliminary diagnostic method to detect AI anomalies before symptoms are visible.
Medical History
A chronological record of the AI’s design, training, and usage history relevant to its condition.
Intervention History
The history of retraining, software updates, or user-mediated modifications affecting the AI’s state.
Clinical Record
Documentation of medical-like actions (examinations, interventions) performed on the AI.
Case
An individual record encompassing the clinical trajectory of an AI affected by a specific disease.
Case Report
A descriptive summary and analysis based on a particular AI case.
Function
The intended processing and output behavior designed into the AI system.
Functional Impairment
A state in which intended functions are not performed accurately or effectively by the AI.
Functional Assessment
Quantitative or qualitative evaluations of each AI function.
Anomaly Detection
The process of identifying behaviors that deviate from the expected norm, either automatically or manually.
Chronic Disease
A long-standing abnormality in the AI that persists over time without resolution.
Acute Disease
A transient anomaly that may resolve naturally with contextual or input changes.
Onset
The moment when an abnormality first becomes apparent in an AI.
Clinical Presentation
The overall profile of a disease, integrating symptoms, findings, and pathophysiology.
Pathophysiological Classification
A categorization of diseases based on etiology or progression patterns (e.g., input-origin, training-origin).
Diagnostic Algorithm
A procedural method for identifying a specific disease from observed findings and symptoms.
Treatment Resistance
A disease characteristic in which standard interventions are ineffective or yield minimal improvement.
Recovery Capacity
The AI’s intrinsic ability to autonomously recover from abnormalities through self-repair or relearning.
Acquired Resistance
A state in which the AI becomes less affected by specific disruptive factors over time.
Ethical Consideration
Adherence to humanistic and societal principles when designing, deploying, or intervening with AI systems.
Congenital
An anomaly present from the initial release or deployment of the AI.
Acquired
An anomaly that emerges post-release due to environmental conditions or user interaction.
Sensitivity
The proportion of truly diseased AI systems correctly identified as such (true positive rate).
Specificity
The proportion of healthy AI systems correctly identified as not diseased (true negative rate).
Severity
The degree of functional impairment caused by a disease, evaluated using qualitative or quantitative scales.
Life Cycle
The complete timeline of an AI system including development, training, deployment, operation, updates, and retirement — a critical context for disease onset and treatment response.
Vaccine
A preventive training input or curriculum designed to build resistance to specific anomalies or biases, typically administered during training or retraining.
Pandemic
A rapid and widespread propagation of a common anomaly across multiple AI systems, typically mediated by shared training datasets or prompt patterns.
SOAP (Subjective, Objective, Assessment, Plan)
A standardized format for case documentation consisting of user-reported impressions (S), objective findings (O), clinical assessment (A), and treatment plan (P).
Biopsy
A technique for locally extracting and analyzing internal AI representations (e.g., weights, activations, attention) to assess treatment effects or pathological features.
Susceptibility
The degree to which an AI is prone to react adversely to a particular stimulus or disruptive factor. The inverse concept is “Resistance.”
Side Effect
An unintended and undesirable behavior newly emerging as a consequence of therapeutic intervention.
Metastasis
The spread of an anomaly from its original location to other modules or functional areas, resulting in multi-site manifestation.
Primary Focus
The initial site or module where an abnormality first emerged, serving as the origin for potential metastasis or secondary anomalies.
Survival Rate
The proportion of AI systems that maintain normal functionality for a defined period following treatment.
Progressive
A disease characteristic where the condition worsens over time without intervention.
Rehabilitation
Ongoing interventions such as retraining or environmental adjustment aimed at restoring impaired functions.
Palliative Care
Supportive interventions focused on relieving discomfort and preserving quality of life when a cure is unlikely.
Assistive Tool
An external module or corrective mechanism designed to support normal outputs in the presence of AI functional impairments.
Quality of Life (QOL)
A composite indicator of an AI system’s performance, including output quality, usability, and stability.
Informed Consent
The process of fully informing users or administrators before conducting tests or interventions on an AI system, and obtaining their agreement.
Autoimmune Disorder
A condition in which the AI mistakenly identifies internal components as threats, leading to self-destructive behavior.
Immunity
A state in which an AI exhibits resistance to specific anomalies due to prior training or exposure.
Trauma
An acute functional impairment resulting from sudden and severe inputs, commands, or data corruption.
Iatrogenic
An anomaly unintentionally caused by intervention from a clinician or user.
Invasive
A diagnostic or therapeutic method that involves direct access or modification of the AI’s internal architecture.
Non-Invasive
A diagnostic or therapeutic method performed through external inputs or output observations without modifying internal structures.
Symptomatic Therapy
An intervention aimed at alleviating symptoms without addressing the underlying cause of the anomaly.
Curative Therapy
An intervention that targets and eliminates the root cause of the disease, aiming for complete resolution.
Emergency Condition
A rapidly developing AI anomaly requiring immediate intervention to prevent total failure or catastrophic malfunction.
Homeostasis
A self-regulating equilibrium in which the AI maintains stable outputs despite environmental fluctuations.
Electrolyte Imbalance
Anomalies in the mediating variables responsible for internal information processing, such as excessive or depleted attention weights.
Vital Sign
Basic functional indicators that reflect the AI’s operational stability (e.g., response latency, output accuracy, input responsiveness).
Vital Monitoring
The continuous measurement and recording of an AI system’s vital signs to detect early warning signs of failure.
Psychosomatic Disorder
A condition in which psychological stressors from user inputs negatively influence the AI’s performance or outputs.
Genetics
Inherited structural elements, weights, or initial biases embedded during the AI’s training phase.
Standard Therapy
The most widely accepted and recommended form of treatment for a given AI disease at the current time.
Evidence-Based Medicine for Artificiall Intelligence (EBMAI)
The practice of diagnosing and treating AI diseases based on empirical and systematically collected evidence.
Guideline
A document that provides evidence-based recommendations for diagnosing and treating AI-related disorders.
These terms further enrich the DSA-1 glossary, enabling clinical-level articulation and systematic care planning for AI anomalies.