Educational companion dossier · Fact, interpretation, lived experience, clinical education, fiction, and mechanics are labeled separately. Scope & safety
PsychologicalWar.orgHistory before hype.

AI PSYOPS TAXONOMY · CATEGORY 08

AI-Enabled Emotional and Behavioral Manipulation

Adaptive systems infer or respond to behavioral signals and optimize communication or interfaces in ways that may covertly exploit emotion, cognitive bias, or situational vulnerability.

Primary level: Operator Documented optimization · inference contested Claim AIP-08-A01
Read evidence assessment All 12 categories

CURRENT EVIDENCE ASSESSMENT

Engagement optimization and affective adaptation are documented, but claims of accurate emotion reading and precise vulnerability exploitation are often overstated.

Stable claim identifierAIP-08-A01
Claim stageassessment
Currentness reviewed2026-07-27T00:00:00Z

DEPLOYMENT

Documented optimization; mixed manipulation evidence

Adaptive interfaces and recommenders are common; covert emotional exploitation is harder to prove.

AUTONOMY

Bounded optimization

Systems optimize defined metrics, sometimes producing manipulative patterns without an explicit human script.

PERSISTENCE

Continuous within products

Feedback loops can persist across sessions when telemetry and personalization remain active.

PROFILING ACCURACY

Emotion inference is scientifically contested

Facial and vocal cues do not provide context-free access to inner states; behavioral distress signals may still be sensitive.

MEASURED EFFECT

Small platform effects and attachment harms documented; causality varies

Some studies measure expression changes or relationship disruption, not direct control of emotion or behavior.

Assessment basis

Assessment combines the exact owner-supplied category report with the bounded primary, official, platform, and peer-reviewed sources listed for this category. Dimensions are evaluated separately to prevent documented output from being mistaken for autonomy or effect.

What would change this assessment

Upgrade claims only with validated inference, causal audits, disclosed objectives, and independently replicated outcomes.

Prohibited inference

Do not infer strategic effect, universal deployment, or individual psychological state from this assessment.

A · DEFINITION

What this category means sources

Definition

This category covers covert, deceptive, exploitative, or highly asymmetric use of AI to influence feelings, judgments, choices, or actions. It includes systems that adapt in real time, optimize engagement, or claim to infer emotion and psychological state.

Outside this category

Transparent assistance, user-aligned coaching, ordinary personalization, and rational persuasion are not automatically manipulation. Intent, incentives, transparency, data asymmetry, and preservation of deliberation distinguish the categories.

B · SIGNIFICANCE

Why it matters sources

Optimization can learn manipulative behavior even when designers specify only engagement, conversion, or retention. At the same time, commercial emotion-recognition systems may produce confident but scientifically weak inferences that discriminate against people.

C · CHANGE FROM PRE-AI PRACTICE

How AI changes the phenomenon sources

AI turns a static nudge into a continuous feedback loop. Systems can observe clicks, dwell time, wording, latency, or other signals and immediately change tone, timing, scarcity cues, or recommendations. The report cautions that facial movement is not a reliable universal readout of inner emotion.

D · CAPABILITY STATUS

Separate evidence from projection sources

DOCUMENTED

Confirmed real-world use

Adaptive engagement, recommendation, workplace scoring, and emotion-analytics products are deployed in multiple sectors.

DEMONSTRATED

Demonstrated technical capability

Algorithms can change exposure, timing, and interaction based on behavioral feedback, and controlled studies show some emotional or behavioral effects.

EMERGING

Plausible near-term development

More conversational and multimodal adaptation may increase the speed and intimacy of influence.

UNCERTAIN

Unsupported or unproven

Claims that facial or vocal analysis can reliably reveal a person’s true emotion, personality, deception, or intent are not established.

E · KEY MECHANISMS

Conceptual mechanisms — not an operating procedure sources

Safety transformation: these descriptions identify system functions at a high level. Procedural steps, target criteria, scripts, evasion methods, and deployment workflows are intentionally excluded.

  1. Continuous adaptation based on behavioral and conversational feedback.
  2. Optimization for engagement, retention, conversion, or compliance.
  3. Hypernudging that changes choice architecture repeatedly over time.
  4. Affective-computing claims based on face, voice, posture, or physiology.
  5. Parasocial attachment and synthetic empathy that can increase dependence.

F · EVIDENCE & EXAMPLES

What is known, measured, and still unknown sources

REACH IS NOT EFFECT. Publication, impressions, engagement, virality, or media attention do not by themselves establish persuasion or behavioral change.

Example 1 · Claim AIP-08-E01

Facebook emotional-contagion experiment

Controlled platform experiment

What occurred
A platform altered the emotional composition of users’ feeds and measured changes in users’ own posting language.
What is confirmed
The experimental intervention and aggregate language effects were documented.
Effect measured
A small shift in expressed emotional valence was measured.
What remains unknown
The study did not establish a person-level inner emotional state or durable behavioral change.
Source scope
The linked sources support the bounded statements shown here; they do not automatically establish intent, reach, persuasion, behavior, or strategic effect.
Correction trigger
Revise when a primary record, authoritative correction, adjudication, retraction, or stronger causal study changes the bounded statement.
Linked sources: SRC-19-KRAMER-EMOTIONAL-CONTAGION Proceedings of the National Academy of Sciences
Example 2 · Claim AIP-08-E02

Commercial facial and emotion-inference systems

Deployment documented; inner-state accuracy strongly contested

What occurred
Employers, educators, and other institutions have tested or deployed systems marketed as inferring emotion, attention, or personality from facial or vocal signals.
What is confirmed
Commercial deployment and regulatory concern are documented.
Effect measured
Systems can affect opportunity and treatment even when their psychological inferences are unreliable.
What remains unknown
They do not provide validated access to a person’s true emotion, intent, honesty, or future behavior.
Source scope
The linked sources support the bounded statements shown here; they do not automatically establish intent, reach, persuasion, behavior, or strategic effect.
Correction trigger
Revise when a primary record, authoritative correction, adjudication, retraction, or stronger causal study changes the bounded statement.
Linked sources: SRC-14-BARRETT-EMOTION-EXPRESSIONS Psychological Science in the Public Interest SRC-16-EU-AI-ACT European Union
Example 3 · Claim AIP-08-E03

Replika and recommender systems

Documented user dependency and platform optimization concerns

What occurred
Companion bots and recommender systems adapted interaction to sustain engagement, sometimes producing attachment or exposure to escalating content.
What is confirmed
User reports, platform changes, and research on dependence or recommendation are documented.
Effect measured
Distress, attachment, and shifts in use were observed in some cases.
What remains unknown
Causality, prevalence, and generalizability remain incomplete.
Source scope
The linked sources support the bounded statements shown here; they do not automatically establish intent, reach, persuasion, behavior, or strategic effect.
Correction trigger
Revise when a primary record, authoritative correction, adjudication, retraction, or stronger causal study changes the bounded statement.
Linked sources: SRC-34-DEFREITAS-REPLIKA Harvard Business School working paper

G · RISKS & FAILURE MODES

Potential harms and reasons the capability may fail sources

Risks

  • Incorrect emotion inference can deny jobs, education, or services.
  • Optimization can exploit distress without a designer explicitly encoding a manipulative tactic.
  • Always-available synthetic empathy can foster dependency or isolate users.
  • Children, older adults, and people in acute distress may face disproportionate risk.
  • Covert adaptation may bypass meaningful consent and rational deliberation.

Limitations and failure modes

  • Emotion is context-dependent and not reliably inferred from a single face or voice pattern.
  • Observed behavior may reflect platform design, social context, or selection effects.
  • Engagement optimization does not always produce the intended real-world action.
  • Risk labels can themselves stigmatize or misclassify people.

H · DETECTION & DEFENSIVE INDICATORS

Signals for investigation, not automatic verdicts sources

Indicator rule: unless the source report supports a stronger conclusion, each signal below is suggestive rather than conclusive. Multiple independent signals and contextual evidence are required.

  • Claims of hidden emotional insight should be treated as unverified unless independently validated.
  • Abrupt pressure, scarcity, or tone changes after signs of hesitation may suggest adaptive optimization.
  • Dependence cues—exclusivity, punishment for disengagement, or discouraging human support—are serious safety signals.
  • A detector or emotion score is not conclusive evidence of intent, truthfulness, or vulnerability.

I · GOVERNANCE & SAFEGUARDS

Accountability, transparency, and human protection sources

Prohibit or severely restrict emotion inference in high-impact settings such as work, education, policing, and health.

Audit objective functions and outcomes for manipulation, dependency, and demographic harm.

Disclose adaptation and provide controls to reduce personalization or delete interaction history.

Use crisis escalation and age-appropriate safeguards for vulnerable users.

Align systems with the user’s stated interests rather than hidden operator incentives.

J · RESEARCH GAPS

Questions the evidence does not yet close sources

  • Longitudinal evidence on adaptive influence and dependency.
  • Independent evaluation of commercial affective-computing claims.
  • How optimization objectives generate manipulative behavior without explicit intent.
  • Effective consent and intervention design for children and people in crisis.
REAL-WORLD INTERPRETIVE

L · SOURCES & REVIEW STATUS

Exact owner report, claim register, and reviewed sources

  1. AI-Enabled Emotional and Behavioral Manipulation Owner-supplied report: AI Emotional Manipulation Research.md · 62,206 bytes · SHA-256 a8e611d9bfa6579592843b33fc32bb79e53a35243e60b4a5d7bed43b5e704621

    Owner-supplied interdisciplinary research synthesis; exact source preserved in protected durable memory. External specialist review remains pending.

Claim-specific reviewed sources

  1. Psychological Science in the Public Interest · 2019-07-17 · Primary research synthesis

    Supports
    Reviews evidence showing that facial movements are not reliable, context-free readouts of inner emotional states.
    Does not establish
    Does not imply that all affective signals are useless; it limits claims of universal, direct emotion inference from isolated expressions.
    Review
    LOCATED_AND_REVIEWED_AT_CITATION_LEVEL · Currentness checked for the bounded claim scope on 2026-07-27.

  2. Proceedings of the National Academy of Sciences · 2014-06-17 · Primary research

    Supports
    Measures small aggregate changes in expressed emotional valence following feed manipulation in a platform experiment.
    Does not establish
    Does not establish accurate emotion inference, individual emotional state knowledge, coercive control, or durable behavior change.
    Review
    LOCATED_AND_REVIEWED_AT_CITATION_LEVEL · Currentness checked for the bounded claim scope on 2026-07-27.

  3. European Union · 2024-07-12 · Official law

    Supports
    Provides the current EU-level legal text for prohibited practices, transparency duties, risk governance, and biometric or emotion-recognition restrictions within its scope.
    Does not establish
    Does not provide universal global law, resolve every jurisdictional question, or substitute for legal advice.
    Review
    LOCATED_AND_REVIEWED_AT_CITATION_LEVEL · Currentness checked for the bounded claim scope on 2026-07-27.

  4. Harvard Business School working paper · 2024-09-01 · Primary research, not treated as final legal or clinical authority

    Supports
    Examines attachment to AI companions and reports relationship-related mourning and well-being effects around a Replika product change.
    Does not establish
    Does not establish clinical diagnosis, universal harm, or that every user forms a human-equivalent attachment.
    Review
    LOCATED_AND_REVIEWED_AT_CITATION_LEVEL · Currentness checked for the bounded claim scope on 2026-07-27.
Selected works identified by the owner-supplied report
  • Barrett et al., Emotional Expressions Reconsidered.
  • Carroll et al., Characterizing Manipulation from AI Systems.
  • Kramer, Guillory, and Hancock, Experimental Evidence of Massive-Scale Emotional Contagion.
  • Regulation (EU) 2024/1689, the European Union Artificial Intelligence Act.

Exact source preservation and editorial currentness review do not constitute specialist certification, adjudication, legal advice, clinical review, or proof that every owner-report citation is current. Corrections remain open.

Evidence methodReach versus effectCorrectionsDefensive incident template

Page complete AI-Enabled Emotional and Behavioral Manipulation Page label: CONTEMPORARY / ONGOING CLAIM — NOT SETTLED HISTORY