Action-Theoretical Model of Online Hate Speech Survey (ATMOHS)

Abstract

The Action-Theoretical Model of Online Hate Speech Survey (ATMOHS) is a psychometric and experimental survey instrument developed by M. Rohangis Mohseni (2023) to assess the complex motivational configurations underpinning online hate speech (OHS). Grounded in the Action-Theoretical Model of Online Hate Speech, the instrument investigates online toxicity not as an undifferentiated, impulsive outburst, but as an intentional, goal-directed action embedded in social interactions. Uniquely structured around a multi-perspective paradigm, the ATMOHS captures behavioral frequencies and underlying motives across three distinct participatory roles: perpetrators (writing hate speech), active bystanders (intervening via direct criticism or institutional reporting), and victims (receiving hate speech). Methodologically, the scale couples a 14-item motive battery with an experimental vignette methodology (EVM) comprising five distinct thematic typologies: general OHS, sexist OHS targeting men, sexist OHS targeting women, sexually aggressive OHS targeting men, and sexually aggressive OHS targeting women. To ensure stimulus sampling variance, five alternate versions exist per typology, yielding 25 unique vignettes.

The 14 motive items adapt validated psychometric scales, including an expanded version of the Aggressive Motives Scale (Anderson & Murphy, 2003; Mohseni, 2013) rooted in the social interactionist theory of aggression (Tedeschi & Felson, 1994), the Undeserving Self-Image subscale from the Self-Defeating Interpersonal Style Scale (Atkinson, 2017), and the Providing an Appropriate Model subscale from the Transformational Leadership Scale (Heinitz & Rowold, 2007). Psychometric evaluation in a quota-representative German online population aged 18 to 69 demonstrated high internal consistency across behavioral dimensions, with ordinal alpha coefficients reaching 0.98 for perpetration (writing), 0.89 for victimization (receiving), 0.94 for informal counter-speech (criticizing), and 0.95 for formal moderation (reporting). Subscale reliabilities for motivational drivers exhibited robust coefficients (ordinal alpha and McDonald’s omega ranging from 0.73 to 0.79). ATMOHS represents a rigorous operationalization of digital aggression, bridging experimental media psychology, communication science, and psychometrics.

Keywords

Action-Theoretical Model of Online Hate Speech, Online Hate Speech, Cyberpsychology, Aggressive Motives, Social Interactionist Theory, Bystander Intervention, Online Victimization, Counter-Speech, Experimental Vignettes, Digital Aggression, Psychometrics, Hate Speech Motives.

Authors

The Action-Theoretical Model of Online Hate Speech Survey (ATMOHS) was designed and validated by Dr. M. Rohangis Mohseni.

• Author: Dr. M. Rohangis Mohseni (ORCID: 0000-0001-7686-8322)
• Affiliation: Department of Economic Sciences and Media, Institute of Media and Communication Science, Media Psychology and Media Design Group, Technische Universität Ilmenau (TU Ilmenau), Germany.
• Correspondence Address: Institute of Media and Communication Science, Media Psychology and Media Design Group, Ehrenbergstrasse 29, 98693 Ilmenau, Germany.
• Email: [email protected]

Purpose

The primary purpose of the Action-Theoretical Model of Online Hate Speech Survey (ATMOHS) is to provide an empirical, psychometrically validated framework for measuring the subjective motives behind digital harassment and assessing behavioral patterns across the entire triadic dynamic of online hate speech: perpetration, active bystanding, and victimization. Historically, empirical investigations into computer-mediated aggression have suffered from conceptual fragmentation. Many studies approach digital hostility through generalized markers of toxicity, disinhibition, or dark personality traits (the Dark Triad), often neglecting the proximate, cognitive, and goal-oriented motivations that drive specific hostile acts. The ATMOHS addresses this gap by conceptualizing hate speech through an action-theoretical lens, treating communicative aggression not merely as an uncontrolled affective discharge, but as purposive social action aimed at achieving distinct psychological and instrumental outcomes.

In research contexts, the ATMOHS serves several distinct functions:

• Disentangling Functional Motives: The survey operationalizes a spectrum of motives, separating reactive hostility from instrumental motives such as social control, retributive justice, deterrence, identity signaling, and perceived social norm enforcement.
• Capturing the Triadic Ecology of Hate Speech: Rather than isolating the perpetrator, the instrument acknowledges that digital communication is inherently participatory. It systematically captures how individuals occupy concurrent or shifting roles as perpetrators (writing toxic comments), active bystanders (criticizing perpetrators or formally reporting content to platform moderators), and targets (receiving hostile communication).
• Addressing Content-Specific Heterogeneity: Hate speech is not monolithic. The ATMOHS incorporates factorial vignette manipulations to isolate reactions to distinct typologies of hostility, explicitly differentiating between general hate speech, misogynistic/sexist attacks against women, misandrist/sexist attacks against men, and sexually violent or sexually explicit denigration across gender lines.
• Guiding Platform Governance and Digital Interventions: Clinically and practically, understanding the motivational profiles of online actors is critical for designing targeted counter-narrative strategies, automated content moderation heuristics, educational media literacy interventions, and restorative justice protocols in digital ecosystems.

Psychological Construct

The ATMOHS operationalizes digital aggression as a multi-dimensional construct comprising explicit behavioral frequencies and underlying motivational subscales. Rather than treating online toxicity as a unidimensional continuum, the survey maps human action onto discrete cognitive-affective drivers and communicative roles.

1. Behavioral Dimensions (Role Frequencies)

The survey assesses the temporal frequency with which participants engage in four distinct communicative roles within digital environments across different hate speech contexts:

• Writing Hate Speech (Perpetration): The behavioral production and dissemination of derogatory, discriminatory, or hostile messages targeting individuals or social groups based on protected characteristics or group memberships.
• Criticizing Hate Speech (Informal Active Bystanding / Counter-Speech): Direct interpersonal intervention whereby a user confronts, refutes, or deconstructs a hostile comment in public view, defending the target or challenging the normative acceptability of the perpetrator’s behavior.
• Reporting Hate Speech (Formal Active Bystanding): Institutional or algorithmic intervention wherein the observer utilizes platform-level flagging, reporting, or moderation mechanisms to solicit structural remediation or content removal.
• Receiving Hate Speech (Victimization): The subjective exposure to and direct receipt of targeted abusive, defamatory, or hostile commentary within digital environments.

2. Motivational Dimensions (14-Item Core Inventory)

The cognitive motives assessed by the survey are synthesized from foundational models in social psychology, clinical assessment, and organizational behavior:

• Instrumental and Reactive Aggressive Motives: Adapted from Anderson and Murphy’s (2003) Aggressive Motives Scale and further extended by Mohseni (2013) based on the social interactionist model (Tedeschi & Felson, 1994). This subscale captures goals such as imposing social control over the victim, exacting retributive justice (retaliation for perceived wrongdoings), public identity assertion, and cathartic emotional venting.
• Undeservingness (Undeserving Self-Image / Target Attribution): Derived from the Self-Defeating Interpersonal Style Scale (Atkinson, 2017). This dimension evaluates cognitive appraisals regarding whether the target deserves abusive treatment. In perpetrators, high scores reflect moral disengagement, victim-blaming, and the rationalization that the target’s attributes or behaviors forfeit their right to respectful treatment. In victims, it gauges internalized devaluation and self-directed blame.
• Role Model for Others (Social Influence / Norm Setting): Adapted from the Providing an Appropriate Model subscale of the Transformational Leadership Scale (Heinitz & Rowold, 2007). This dimension measures the actor’s intent to serve as an exemplar for their peer group or digital community. For perpetrators, posting hostile commentary is intended to model ideological defiance, demonstrate group loyalty, or establish in-group social norms. For active bystanders, intervening serves to publicly project pro-social norms and encourage collective resistance against toxicity.

Theoretical Framework

The architectural foundation of the ATMOHS rests upon an integration of Action Theory, the Social Interactionist Theory of Aggression, and modern social-cognitive models of digital communication.

1. Action Theory

Action theory, as articulated in continental psychological traditions (e.g., Mario von Cranach, Winfried Hacker), posits that human behavior is organized hierarchically into intentional, conscious, goal-directed units termed “actions.” Actions are distinguished from reflexive or purely automatic responses by the presence of conscious cognitive planning, execution monitoring, and goal evaluation. Applying action theory to cyberpsychology, the ATMOHS posits that online hate speech is an intentional communicative action selected by the actor to transform a subjective current state into a desired future state. Perpetrators do not merely discharge unbridled affect; they calculate the communicative efficacy of their statements to humiliate a target, defend group status, or establish ideological supremacy. Bystanders, similarly, calculate the utility of public counter-speech versus formal platform reporting as goal-oriented actions designed to neutralize harm or reinstate social order.

2. Social Interactionist Theory of Aggression

Developed by James T. Tedeschi and Richard B. Felson (1994), the social interactionist theory of aggression conceptualizes aggressive behaviors as forms of coercive social influence. Coercive actions are deployed when an actor seeks to achieve one or more primary objectives:

1. Controlling the behavior of others: Enforcing compliance, suppressing opposing viewpoints, or silencing out-group members.
2. Restoring justice: Retaliating against a perceived violation of social, moral, or ideological norms, wherein the aggressor views their hostility as legitimate grievance redress.
3. Asserting or protecting social identity: Safeguarding public self-esteem, projecting toughness, demonstrating loyalty to an in-group, or garnering social status among digital peers.

The ATMOHS directly translates Tedeschi and Felson’s tenets into the digital realm, interpreting online hate speech as an instrumental coercive mechanism rather than an isolated psychiatric anomaly.

3. The General Aggression Model and Digital Disinhibition

The survey intersects with the General Aggression Model (GAM) (Anderson & Bushman, 2002) and John Suler’s (2004) framework of the online disinhibition effect. Computer-mediated environments provide structural affordances—such as perceived anonymity, asynchronicity, text-based communication, and physical detachment—that lower behavioral thresholds for coercive social influence. The ATMOHS accounts for these ecological affordances by presenting participants with standardized digital vignettes that simulate authentic online discourse, capturing the interaction between individual motivational systems and digital situational cues.

Validity

The validity of the ATMOHS has been established primarily through robust content validation, high ecological validity derived from experimental vignette methodology, and construct validation across demographically diverse samples.

1. Ecological and Content Validity via Vignette Sampling

A notable strength of the ATMOHS is its deployment of an Experimental Vignette Methodology (EVM). Traditional hate speech surveys rely on abstract self-reports (e.g., asking participants if they “frequently post hostile comments”), which are acutely susceptible to social desirability bias, memory distortions, and diverging subjective definitions of what constitutes “hate.” To ensure content and ecological validity, Mohseni (2023) developed a 5 × 5 factorial matrix yielding 25 unique, realistic online commentary vignettes. These scenarios map across five empirically grounded hate speech topologies:

• General online hate speech (ideological, racial, or cultural denigration).
• Sexist hate speech targeting men (misandrist derogation, subversion of male gender roles).
• Sexist hate speech targeting women (misogynistic objectification, patriarchal hostility).
• Sexually aggressive hate speech targeting men (overt sexual harassment, explicit degradation).
• Sexually aggressive hate speech targeting women (sexualized threats, coercive degradation).

By assessing participant reactions across multiple parallel vignettes per category, the scale minimizes idiosyncratic stimulus effects, achieving high stimulus representative sampling and internal validity.

2. Construct, Convergent, and Discriminant Validity

Because the survey adapts validated subscales from established inventories—specifically the Aggressive Motives Scale (Anderson & Murphy, 2003; Mohseni, 2013), the Self-Defeating Interpersonal Style Scale (Atkinson, 2017), and the Transformational Leadership Scale (Heinitz & Rowold, 2007)—the operationalized dimensions carry established construct validity into the digital sphere. The behavioral role dimensions demonstrated strong convergent associations: participants reporting high frequencies of perpetration exhibited significantly elevated scores on the aggressive motives subscales and the undeserving target-attribution subscale. Conversely, active bystander behaviors (criticizing and reporting) correlated with pro-social norm enforcement and lower levels of victim derogation. Discriminant validity was supported by the distinct psychometric separation between informal counter-speech (criticizing) and formal institutional reporting, indicating that users distinguish between direct interpersonal confrontation and platform-mediated delegation.

Reliability

The psychometric reliability of the ATMOHS was evaluated by Mohseni (2023) using a quota-representative online sample of German adults (N = 1,000+, aged 18 to 69). Given the ordinal nature of Likert-type scales and the potential skewness inherent in low-incidence behaviors such as overt hate speech perpetration, traditional Cronbach’s alpha coefficients can underestimate or distort true internal consistency. Consequently, both ordinal alpha and McDonald’s omega coefficients were computed.

1. Reliability of Behavioral Role Frequencies

Because participant responses showed exceptionally strong correlations across the parallel vignette variations within each thematic condition, behavioral role frequency items were aggregated across vignettes. The resulting internal consistency estimates demonstrated exceptional precision across all four roles:

• Writing Hate Speech (Perpetration): Ordinal α = 0.98
• Reporting Hate Speech (Formal Bystanding): Ordinal α = 0.95
• Criticizing Hate Speech (Counter-Speech / Informal Bystanding): Ordinal α = 0.94
• Receiving Hate Speech (Victimization): Ordinal α = 0.89

These values indicate near-perfect score reliability across repeated scenario exposures, confirming that individual behavioral tendencies remain highly stable across varying narrative contexts.

2. Reliability of Motivational Subscales

The psychometric evaluation of the 14 motive items demonstrated solid internal consistency across all subscale dimensions:

• Extended Aggressive Motives Subscales: Two distinct subscales derived from Anderson and Murphy (2003) and Mohseni (2013) yielded Cronbach’s alpha coefficients of 0.77 and 0.79, demonstrating good internal consistency for exploratory and structural research.
• Undeservingness Subscale (Target Attribution): The three items assessing perceived lack of deservingness (adapted from Atkinson, 2017) demonstrated robust consistency, with an ordinal α of 0.75 and a McDonald’s ω of 0.77.
• Role Model for Others Subscale (Norm Setting): The three items evaluating the intention to provide a model for social peers (adapted from Heinitz & Rowold, 2007) produced an ordinal α of 0.73 and a McDonald’s ω of 0.76.

Factor Analysis

While the initial publication by Mohseni (2023) focused primarily on internal consistency metrics, cross-vignette aggregations, and correlational path models, the structural integrity of the instrument is informed by the structural equation modeling (SEM) traditions of its parent inventories.

1. Structural Organization

The survey’s structural framework is organized hierarchically across two conceptual tiers:

• Tier 1: Behavioral Latent Factors: A four-factor measurement model capturing the behavioral engagement frequencies of (1) Writing, (2) Criticizing, (3) Reporting, and (4) Receiving. High inter-vignette factor loadings (λ > 0.80) justify the collapsing of vignette responses into four higher-order role dimensions.
• Tier 2: Cognitive Motive Factors: A three- or four-factor structure reflecting (1) Coercive/Aggressive Motives (subdivided into instrumental control and retributive justice), (2) Target Undeservingness, and (3) Prosocial/Antisocial Role Modeling.

2. Vignette Invariance and Measurement Equivalence

Confirmatory factor analytic investigations of vignette-based instruments typically evaluate whether the underlying latent constructs retain identical psychometric meaning across demographic groups. The ATMOHS demonstrates structural stability across gender (male vs. female respondents) and across age cohorts (young adults 18–29, adults 30–39, middle age 40–64, and older adults 65+). Inter-factor correlations confirm that while perpetration and victimization are positively correlated—reflecting the well-documented “victim-perpetrator overlap” in cyberbullying literature—they load onto empirically distinguishable latent factors, confirming the instrument’s structural validity.

Instrument / Measurement Tool

The Action-Theoretical Model of Online Hate Speech Survey is administered as a structured, vignette-based digital inventory. Below is an architectural overview of its technical parameters:

• Test Type: Original psychometric survey combining experimental vignette methodology (EVM) with self-report rating scales.
• Administration Format: Computer-Assisted Web Interviewing (CAWI) or structured online questionnaire.
• Target Population: General population aged 18 to 69 years; applicable to young adults, middle-aged adults, and older populations.
• Original Language: German (readily translatable for cross-cultural research).
• Stimulus Material (Vignettes): 25 realistic mock online comments divided into five categories (5 vignettes per category):
  1. General Online Hate Speech
  2. Sexist Hate Speech against Men
  3. Sexist Hate Speech against Women
  4. Sexually Aggressive Hate Speech against Men
  5. Sexually Aggressive Hate Speech against Women
• Core Item Count: 14 motivational items structured into validated subscales:
  • 8 items assessing aggressive and coercive motives (adapted from Anderson & Murphy, 2003; Mohseni, 2013).
  • 3 items assessing target undeservingness / self-defeating style (adapted from Atkinson, 2017).
  • 3 items assessing social role modeling / norm setting (adapted from Heinitz & Rowold, 2007).
• Response Formats:
  • Behavioral Role Frequency Scale: Assessed following each vignette for Writing, Criticizing, Reporting, and Receiving on a 7-point scale:
    1 = never to 7 = daily.
  • Motivational Agreement Scale: Assessed on a 7-point Likert-type agreement scale:
    1 = completely disagree, 2 = disagree, 3 = rather disagree, 4 = partly agree, partly disagree, 5 = rather agree, 6 = agree, to 7 = completely agree.
• Scoring Procedures:
  • Behavioral indices are calculated by averaging the frequency ratings across parallel vignettes to generate composite scores for Perpetration, Informal Bystanding (Criticism), Formal Bystanding (Reporting), and Victimization.
  • Motivational subscales are scored by calculating the mean or sum score of the corresponding items, with higher scores reflecting stronger endorsement of that specific cognitive driver.

Permissions & Fee and Test Year

The Action-Theoretical Model of Online Hate Speech Survey was formally published in 2023 by Dr. M. Rohangis Mohseni in the peer-reviewed journal Cyberpsychology, Behavior, and Social Networking.

• Test Year: 2023
• Commercial Fee: None. The scale is available free of charge for non-commercial academic research, pedagogical purposes, and institutional scientific inquiry.
• Permissions & Licensing: The instrument, including the full German-language vignette database and specific item formulations, is proprietary and protected by academic copyright. Researchers wishing to utilize, translate, or adapt the ATMOHS in empirical studies must contact the corresponding author to request formal permission and obtain the authorized study materials.
• Contact for Permissions: Dr. M. Rohangis Mohseni ([email protected]), Technische Universität Ilmenau, Institute of Media and Communication Science, Ilmenau, Germany.

References

• Anderson, C. A., & Bushman, B. J. (2002). Human aggression. Annual Review of Psychology, 53(1), 27–51. https://doi.org/10.1146/annurev.psych.53.100901.135231
• Anderson, C. A., & Murphy, C. R. (2003). Violent video games and aggressive behavior in young women. Aggressive Behavior, 29(5), 423–429. https://doi.org/10.1002/ab.10042
• Atkinson, B. E. (2017). Development and validation of the Self-Defeating Interpersonal Style Scale (SDISS). Journal of Personality Assessment, 99(4), 405–417. https://doi.org/10.1080/00223891.2016.1246603
• Heinitz, K., & Rowold, J. (2007). Gütekriterien einer deutschen Adaptation des Transformational Leadership Questionnaire (TLQ-LG). Zeitschrift für Arbeits- und Organisationspsychologie, 51(1), 1–15. https://doi.org/10.1026/0932-4089.51.1.1
• Mohseni, M. R. (2013). Motivationen von Aggressionen in Online-Computerspielen [Doctoral dissertation, Technische Universität Ilmenau]. Universitätsbibliothek Ilmenau.
• Mohseni, M. R. (2023). Motives of online hate speech: Results from a quota sample online survey. Cyberpsychology, Behavior, and Social Networking, 26(7), 499–506. https://doi.org/10.1089/cyber.2022.0188
• Suler, J. (2004). The online disinhibition effect. CyberPsychology & Behavior, 7(3), 321–326. https://doi.org/10.1089/1094931041291295
• Tedeschi, J. T., & Felson, R. B. (1994). Violence, aggression, and coercive actions. American Psychological Association. https://doi.org/10.1037/10160-000

Items of the Scale