Media and Technology Psychology

Abstract. Media and Technology Psychology investigates how communication environments, media representations, networked infrastructures, interactive interfaces, and computationally selected information influence attention, memory, judgment, emotion, identity, behavior, social relationships, and collective experience. This monograph reconstructs the field from mass-communication research and computer-mediated interaction through contemporary studies of social platforms, mobile media, recommender systems, and human–artificial relationships. Its method is a critical integration of foundational psychological theories, original experiments, observational and longitudinal studies, replication research, systematic reviews, and recent algorithmic field experiments. The central conclusion is that psychological outcomes depend on the conjunction of content, design affordances, individual characteristics, social context, and measurable interaction patterns. Evidence for a single uniform effect of screen time is weak, while evidence for particular mechanisms and situated interventions is more informative. The independent contribution is the Evidence–Affordance–Agency Matrix, an operational framework that joins exposure architecture to cognitive processes, relational consequences, and the quality of causal inference. It generates researchable hypotheses concerning algorithmic selection, agency, interpersonal reciprocity, and the continuity of digital experience across human and artificial systems.

Keywords: Media psychology; technology psychology; media effects; digital behavior; computer-mediated communication; social media; affordances; attention; social comparison; parasocial interaction; misinformation; media literacy; adolescent development; recommender systems; human–artificial interaction; causal inference; Artificial Era.

The Scientific Object and Boundaries of the Field

Media and Technology Psychology studies psychologically consequential interaction with mediated symbols and technical systems. A medium can distribute representations, organize encounters, shape response options, record behavior, and feed those records into future selections. Technologies therefore operate as conditions of exposure and action. Psychological analysis must specify the perceiving or acting person, the relevant media environment, the concrete psychological process, and the outcome that would count as evidence. A claim that “media changes people” becomes scientific when the medium, mechanism, comparison condition, time interval, and measurement are explicit. The field intersects with communication, cognitive psychology, developmental science, social psychology, learning theory, human-computer interaction, political psychology, and computational social science. Valkenburg, Peter, and Walther (2016) survey this plural intellectual inheritance.

The unit of analysis can be an episode of viewing, a conversation, an extended user trajectory, a social network, or a platform-mediated public. These levels cannot be substituted for one another. A laboratory response to a news frame reveals something about short-term judgment; a year of behavioral logs may describe habitual interaction without determining the causes of those habits. Measures of exposure differ from measures of attention, endorsement, recall, sharing, and long-term behavioral change. A complete theory specifies transitions among these variables. A platform is simultaneously an interface, social setting, information-ranking system, and institution; separating those components produces more precise experimental tests.

Historical Formation: From Mass Media to Interactive Networks

The field emerged through several partially independent research lineages. Studies of persuasion, broadcasting, social learning, public attention, and audience motivations formed a foundation for media-effects research before networked platforms existed. McCombs and Shaw (1972) investigated correspondence between media emphasis and public issue salience; their agenda-setting analysis remains central because it distinguishes what receives attention from what individuals believe about that object. Katz, Blumler, and Gurevitch (1973) articulated a uses-and-gratifications orientation in which audiences select media to pursue goals. The two traditions invite different causal models: exposure can be structured by institutions, while selective use also reflects human motives.

Horton and Wohl (1956) described parasocial interaction as an apparent face-to-face relation with media personae; contemporary creator economies and conversational artificial systems revive questions about perceived reciprocity and interpersonal meaning. Bandura (2001) located symbolic communication within social-cognitive processes including observational learning and perceived agency. Internet communication later added persistent profiles, searchable social histories, interactive feedback, and many-to-many distribution. Personalization introduced selection that changes as people interact. These transitions alter the research object: contemporary exposure is partly a consequence of earlier behavior, and the environment adapts to that behavior. Historical continuity is therefore found in mechanisms such as attention, imitation, affiliation, and social inference, alongside the newly dynamic architecture of delivery.

Mass Communication, Agenda Setting, and Salience

Agenda-setting theory concerns the prominence of issues in public and individual attention. If a topic appears frequently or prominently, it may become easier to recall and more salient in judgments of collective importance. The classic study of the 1968 U.S. election by McCombs and Shaw (1972) provided influential observational evidence for this association. Its interpretation requires separating selection into media and political preferences from media-induced changes in perceived importance. Modern experiments can manipulate ordering, position, or the share of an issue in a feed while measuring subsequent importance ratings. Observational correlations between media and public agendas provide an initial signal, but causal attribution depends on temporal and experimental constraints.

Salience is also related to cognitive accessibility. Tversky and Kahneman (1973) demonstrated how ease of recalling instances can shape judgments of frequency and probability. Repetition, vividness, and recency can influence availability without corresponding changes in the underlying world. Contemporary feeds intensify the importance of this distinction because ranking and repetition create highly unequal encounter probabilities. A rigorous study needs a record of actual impressions and remembered impressions, since the person may see material without encoding it and may recall material seen elsewhere. Salience is a distinct outcome from persuasion, trust, and action.

Audience Motives, Selective Exposure, and Gratifications

Media choice is guided by purposes that include information seeking, connection, diversion, self-expression, reassurance, status negotiation, and identity exploration. The uses-and-gratifications framework corrects a purely broadcast-centered picture by treating audience selection as an explanatory variable. However, motivation is difficult to infer from use alone. A person may open an application intending to contact a friend and remain because of an engaging recommendation. A self-reported motive elicited after a session can rationalize behavior that arose through prompts or habit. Researchers should compare pre-use goals, observed actions, and post-use interpretation.

Selective exposure produces reciprocal relationships between attitudes and media environments. People select content matching their preferences; recommendation systems learn from that selection; the resulting feed may reinforce those preferences. This loop combines individual agency and platform affordances. To estimate it, researchers must observe what content was available but ignored, what was served but skipped, and what was genuinely consumed. Bakshy, Messing, and Adamic (2015) analyzed exposure to cross-cutting political news in a large Facebook dataset, illustrating the need to distinguish network composition, ranking, and individual choice. Findings from one platform and period cannot be treated as universal laws of information selection.

Social Cognitive Theory, Modeling, and Symbolic Influence

Social cognitive theory places learning and behavior inside interactions among personal factors, behavior, and environmental conditions. Bandura's analysis of mass communication (2001) describes direct symbolic influence and socially mediated influence involving networks and community incentives. A media depiction may furnish models of action and imagined consequences; whether a viewer reproduces the behavior also depends on perceived relevance, social reinforcement, ability, and opportunity. The distinction explains why observing content and imitating content are separate research outcomes.

Digital platforms add public counts, comments, and repeated feedback to symbolic models. A visible behavior can appear common or rewarded because a ranking system makes it prominent. A plausible psychological pathway moves from selective visibility to perceived descriptive norms, anticipated rewards, intention, and action. This is a hypothesis that requires separate measures at each stage, ideally with random assignment of signal prominence. Studies that only measure self-reported platform use and later conduct cannot identify whether content, social feedback, or prior disposition produced the relationship. Mechanistic models offer the greatest value when they invite falsifiable mediation tests rather than merely giving a theoretical name to observed correlation.

Computer-Mediated Communication and Relational Affordances

Technology-mediated conversation introduces choices about synchrony, permanence, visibility, editing, anonymity, identity presentation, and the audience. These affordances can alter how speakers formulate messages and how recipients interpret them. A composed personal message entails different relational information from a one-click reaction, even when both are counted as engagement. Burke and Kraut (2016) connected Facebook logs to well-being reports in a three-wave panel of 1,910 participants; targeted, composed communication from strong ties was associated with improved well-being, while generic broadcasts and low-effort feedback were not associated with comparable gains. The observational design supports a differentiated relational interpretation without proving that the message caused the improvement.

The practical construct is communicative specificity: who addressed whom, with how much effort, under what visibility constraints, and with what interpersonal history. Aggregate counts flatten these differences. Design research should therefore measure reciprocity, partner identity, conversational responsiveness, and satisfaction beyond frequency of sending or receiving. Communication outcomes depend also on accessibility, language, disability, culture, and whether a channel enables participation that would otherwise be difficult. Technical affordances acquire psychological significance through use within relationships.

Parasocial Experience and Mediated Social Presence

Parasocial interaction refers to the experience of engagement with a media figure as though participating in a familiar social relationship. The formulation of Horton and Wohl (1956) concerned broadcast personae and the intimacy implied by repeated encounters. Current media environments make the object more complex: creators may answer selected followers, recommendation systems repeatedly deliver familiar faces, and conversational agents can provide contingent replies. A response that appears personally directed may have been produced for many viewers or generated automatically. The psychological target is a person's interpretation of relational signals, not an assumption about the other party's internal experience.

Research can separate perceived social presence, felt closeness, imagined reciprocity, actual response contingency, and the effect of engagement on offline social life. Existing personal attachments and social isolation may shape both content selection and parasocial experiences, producing selection bias in cross-sectional comparisons. Longitudinal designs can estimate whether growing contact predicts changes in perceived relationship quality; experiments that vary response personalization or disclosure can test immediate appraisals. Ethical questions include manipulation of perceived intimacy, commercial persuasion through trusted personae, and the right to understand when interaction is automated. The concept is a bridge toward rigorous studies of human–artificial relational psychology.

Attention, Distraction, and Media Multitasking

Attention research investigates how people allocate limited processing resources across competing stimuli, tasks, and goals. Media multitasking refers to use of several streams concurrently or in rapid alternation. In an influential study, Ophir, Nass, and Wagner (2009) found that heavy media multitaskers performed less well on some cognitive-control tasks than lighter users. The study examined between-person associations; frequent multitasking was not randomized as a sustained life intervention. The association therefore cannot establish that media exposure caused weaker control. Self-report indices also depend on memory and interpretations of simultaneity.

Subsequent research changed the evidentiary picture. Two replication studies and a meta-analysis published in 2017 reported only five significant outcomes in the predicted direction across fourteen replication tests, and a weak aggregate association that became nonsignificant after correction for small-study effects. This makes the case exemplary for theoretical revision: a plausible mechanism and an original significant result cannot substitute for replicated measurement. Researchers should distinguish interruption frequency, voluntary task switching, environmental distractors, notification timing, and the cognitive requirements of the primary activity. A useful outcome is performance cost under defined task demands, rather than a generalized label of impaired attention.

Memory, Cognitive Offloading, and Information Retrieval

Digital systems allow users to store intentions, facts, plans, and personal histories externally. Cognitive offloading transfers part of a task's representational burden to tools such as search engines, calendars, maps, notes, and conversational systems. Its consequences depend on whether the target is recall of content, recall of a location, successful action, or learning that persists after assistance ends. Faster retrieval can improve performance in a supported environment even when unaided retention differs. These are distinct outcomes, so a research design should specify which one matters.

A credible experiment might randomly assign learners to internal rehearsal, structured external notes, and unstructured search, test immediate comprehension, then test delayed unaided recall and the ability to find information with tools. The analysis should include preexisting expertise, quality of sources, and effort spent verifying responses. Theoretical claims of inevitable dependence or automatic cognitive enhancement require evidence that extends beyond short-term task success. Offloading is also social: people remember through other people, institutions, and shared archives. In the Artificial Era, externally supported memory becomes relevant to continuity of projects and relationships. Psychological study can directly observe retrieval behavior and reliance patterns without presupposing that an artificial system has human-like autobiographical memory.

Emotion, Arousal, Social Comparison, and Self-Evaluation

Emotional responses to media vary with content, personal goals, prior experience, and surrounding relationships. Social comparison may occur when people encounter selectively presented achievements, appearance, or status. The affective meaning of the same representation can differ according to identification, aspiration, threat, and the perceived attainability of the comparison target. A laboratory contrast between upward comparison and supportive communication can test immediate affect, whereas ecological momentary assessments reveal when these experiences occur during ordinary life. Self-report of “social media time” obscures the content and context of each episode.

Longitudinal research linking activity logs with repeated well-being measurement offers greater temporal information than retrospective surveys but remains vulnerable to unmeasured changes in circumstances. The Burke–Kraut study is particularly important because it distinguishes targeted relational communication from broadcast exposure. Media environments also provide opportunities for emotional support, shared celebration, learning, and identity affirmation. Therefore, an explanation of risk or benefit should identify the proximate mechanism rather than deducing an effect from technology membership. Researchers can improve inference by pairing moment-level exposure descriptions with affect measures before and after encounters and by preregistering which comparisons are theoretically consequential.

Development, Adolescence, and Individual Susceptibility

Developmental consequences of digital media must be studied relative to capacities, needs, and contexts that change with age. Younger children, adolescents, emerging adults, and older adults face different tasks of learning, autonomy, identity, social coordination, and care. Nesi, Choukas-Bradley, and Prinstein (2018) proposed a transformation framework for adolescent peer relations that emphasizes altered frequency, immediacy, permanence, and audience size of social exchanges. Such transformations can intensify both opportunities and social pressures. Applying an adult effect estimate to all adolescents ignores variation in maturity, social position, and surrounding support.

Odgers and Jensen (2020) reviewed adolescent mental health research and called for more precise measurement and attention to individual differences. Developmental inference requires age-specific longitudinal designs, developmental baselines, and sampling beyond affluent or English-speaking populations. Parents' mediation, peer norms, school practices, and local connectivity can alter the opportunity structure. Clinical samples also need deliberate inclusion: research focused on general community samples may not describe young people who already have severe symptoms or disabilities. Effects of particular interactions should be interpreted in relation to social experiences that would otherwise have occurred.

Screen Time as Exposure: What a Number Can and Cannot Mean

Total duration is an accessible measurement but a poor substitute for the content, purpose, timing, and social organization of use. Two participants can report the same number of hours while one maintains relationships, studies, or creates art and the other encounters distressing material or loses sleep. Duration is also affected by school demands, employment, device sharing, accessibility needs, and socioeconomic resources. Analyses treating all minutes as interchangeable create exposure misclassification. Researchers need to describe the device, service, task, and collection method, and distinguish self-reported estimates from actual logs.

Orben and Przybylski (2019) used specification-curve analysis across three large adolescent datasets totaling 355,358 participants. They reported a small negative association between digital technology use and well-being, explaining at most 0.4 percent of its variation under their analyses. This number describes association in those datasets and specifications; it is not a universal estimate for every platform, group, or content category. A superior exposure model records meaningful activities and their timing, including night use, incoming alerts, interpersonal conflict, supportive conversations, and personally relevant tasks. Precise intervention targets arise from the pathways that explain a particular outcome.

Digital Well-Being: Randomized Deactivation and Reduction

Randomized experiments can estimate the consequences of restricting platform access under defined circumstances. Hunt and colleagues (2018) randomized 143 university students to reduce daily use of Facebook, Instagram, and Snapchat to ten minutes each or to continue usual use for three weeks after baseline monitoring. The restriction group reported greater reductions in loneliness and depressive symptoms relative to the control group. The small, university-specific, short-term study illuminates a possible intervention effect, with uncertain extension to other ages, services, durations, or participants with diagnosed conditions.

Allcott, Braghieri, Eichmeyer, and Gentzkow (2020) randomized Facebook deactivation before the 2018 U.S. midterm election. They found increased subjective well-being and offline activity, reduced news knowledge and political polarization, and persistent changes in later Facebook use. These jointly positive and negative effects show why well-being cannot serve as the only evaluation criterion. Randomized restriction removes many experiences at once: comparison, distraction, informational access, and social connection. Experiments that manipulate one affordance or type of content can better identify which mechanism matters. Deactivation studies should record substitution: individuals may replace platform time with television, interpersonal contact, work, or another application, each influencing observed outcomes.

Adolescent Mental Health and Clinical Relevance

Claims about mental health require differentiation between self-reported distress, questionnaire scores, clinically established disorders, and changes in functioning. A cross-sectional relationship between use and depressive symptoms has several plausible causal structures: distress may prompt greater use; harmful encounters may elevate distress; third variables such as social adversity may affect both; or effects may operate in both directions. Fassi and colleagues (2024) synthesized observational studies of social media use and internalizing symptoms in adolescents, including clinical and community participants. The published review covered 143 studies and more than one million adolescents; only a minority of studies concerned clinical samples. A correction issued in July 2026 changed some reported estimates, underscoring the importance of reading the corrected version rather than reproducing unverified historical coefficients.

Across this literature, average associations and high between-study heterogeneity warrant carefully delimited conclusions. Clinical relevance is not determined solely by statistical significance in very large pooled samples. Studies should report validated symptom scales, baseline severity, impairment, follow-up duration, exposure ascertainment, attrition, and contextual stressors. The empirical findings support the value of attentive, context-specific assessment; they do not support diagnosing an individual based on media use alone. Future interventions must evaluate both possible reductions in harmful exposure and the potential loss of social support or access to important resources.

Platform Design, Reinforcement, and Habitual Engagement

Interface design shapes the costs and rewards of continuing or interrupting activity. Autoplay, infinite scrolling, notifications, variable content ordering, and social feedback can change the frequency of encounters and the salience of cues. The psychology of learning provides concepts for testing their effects: conditioned cue responding, reinforcement schedules, goal-directed choice, and the transition from deliberate to habitual action. Nevertheless, observing repeated checking does not establish that a specific interface has created a clinical addiction or eliminated agency. Habit strength, expected reward, time pressure, and opportunity costs should be measured separately.

The relevant experimental contrast is between interface configurations with the same informational and social purpose but different prompts or stopping cues. Outcome measures include number of unplanned returns, disengagement latency, self-reported intention before opening, task interruption costs, and later satisfaction. The distinction between high engagement and high welfare is fundamental: a design optimized for clicks can produce outcomes that users would not choose under reflective conditions, while repeated use may also represent valuable work or relationships. Research should examine whether short-term engagement metrics predict learning, autonomy, safety, or durable satisfaction in representative populations.

Algorithmic Curation and Adaptive Exposure

Recommendation systems influence which content becomes available, when it appears, and how frequently similar content returns. Unlike a fixed television schedule, a personalized feed updates in response to behavior and thereby creates a feedback loop between user preferences and the environment. Important variables include the candidate pool, ranking objective, user controls, reshares, network connections, and exploration of unfamiliar material. A chronological feed is not a neutral counterfactual for all purposes because it also reflects the user's chosen network; nevertheless, it is a useful experimental alternative to engagement-optimized ordering.

Randomized field experiments clarify that platform effects are contingent. In an experiment with U.S. X users, Gauthier, Hodler, Widmer, and Zhuravskaya (2026) reported that switching from a chronological to an algorithmic feed increased engagement and shifted several political opinions in a more conservative direction. Turning the ranking system off for prior algorithm users did not produce a symmetric reversal; affective polarization and party identification were not significantly changed by either switch. The authors linked the asymmetry partly to follow choices persisting after treatment. The conclusion applies to the studied platform, period, participants, and measured outcomes. Algorithmic mediation must be evaluated as a sequence of exposures and durable network adjustments, not solely as an instantaneous change in ranking.

Experimental Algorithm Redesign and Social Norm Perception

Digital environments communicate apparent social norms through the visibility of posts and users. A few highly active or amplified accounts can generate a misleading impression of how widely a view is held. In a 2026 registered report, Brady and colleagues randomized 2,000 participants to alternative feed-ranking procedures for eight weeks around the 2024 U.S. presidential election. Engagement-based ranking amplified intergroup, moralized, emotionally charged, and toxic content relative to reverse-chronological ordering. A designed alternative that reduced the influence of extreme users improved some measures of perceived norms while maintaining comparable platform enjoyment.

This experiment adds an important methodological step: the researcher controls the ranking function and can distinguish objective exposure from people's inferences about other people. It also reveals that shifts in displayed content need not translate to equally large changes in self-reported behavior. Scientific inference requires identifying the measured norm concept, the treatment duration, and the comparison condition. A perception of animosity is distinct from personal animosity; changing a perceived norm is distinct from changing civic conduct. Replications across diverse societies, languages, political settings, and platforms are necessary before generalizing an interface effect to public life overall.

Political Information, Polarization, and Democratic Cognition

Research on digital political communication examines issue exposure, belief accuracy, interpersonal hostility, civic engagement, and political action. These outcomes frequently diverge. The Facebook and Instagram deactivation experiment reported by Allcott and colleagues (2024) involved randomized account deactivation for six weeks before the 2020 U.S. election among subsets of 19,857 Facebook users and 15,585 Instagram users. The study reported less political participation, largely through online activity, and estimates near zero for several polarization and electoral outcomes. A 2026 correction updated the competing-interest statement; the correction is part of the citation history and should be considered by readers evaluating the work.

These findings should be set beside the platform-specific 2026 X-feed experiment. Different platforms, interventions, and historical periods can produce different outcomes. Research should avoid treating the absence of an estimated average effect in one study as proof that feed design never matters, or treating a significant shift in one opinion as a general transformation of political identity. Relevant mechanisms include perceived norm prevalence, selective exposure, elite cues, interpersonal discussion, and trust in institutional reporting. Studies ought to report how data access and platform cooperation constrain independence and replicability.

Misinformation, Truth Discernment, and Sharing Decisions

Accuracy judgments and sharing choices are psychologically distinct. A participant can recognize that a statement is doubtful yet share it for entertainment, affiliation, outrage, or public discussion. Researchers must therefore distinguish belief, confidence, perceived credibility, intention to share, and observed sharing. Guess, Nagler, and Tucker (2019) investigated dissemination of fabricated news on Facebook during a U.S. election context, demonstrating how behavior can be studied rather than inferred from exposure alone. Prevalence estimates are sensitive to the definition of misleading material, data visibility, and platform period.

Pennycook and colleagues (2020) found that prompting attention to accuracy improved discernment in intended sharing of COVID-19 headlines in online experiments. A 2022 synthesis by Pennycook and Rand combined twenty related experiments with 26,863 participants, supporting a replicable average effect for accuracy prompts within the tested procedures. Intentions elicited in a survey remain distinct from platform behavior. This literature demonstrates an actionable cognitive mechanism—attention directed toward veracity—while leaving questions about durability, hostile actors, language diversity, and real-world implementation. Researchers should test whether an intervention reduces false sharing without reducing valuable legitimate sharing or discussion.

Evidence Integrity, Media Literacy, and Intervention Design

Media literacy includes source assessment, recognition of persuasive intent, understanding of platform ranking, and reflective decisions about sharing or relying on information. Knowledge and behavior may diverge: a person can correctly describe how recommendations work yet still respond to content under time pressure. An effective intervention specifies its target, such as improved discrimination between reliable and unreliable claims, and assesses transfer to new material rather than only performance on taught examples. Evidence should be compared across immediate posttests, delayed assessments, and actual behavior where feasible.

Accuracy reminders are a particularly instructive intervention class because experimental evidence specifies a narrow manipulation and a corresponding judgment outcome. However, an educational program or interface label can work through several mechanisms, including trust, attention, confusion, and social signaling. A careful evaluation includes negative controls, preregistered outcome definitions, and tests for unequal effects by prior knowledge or political identity. The goal is informed agency: users should be able to understand how the media environment shapes exposure and choose practices serving their legitimate purposes. Better cognitive decisions cannot be equated automatically with greater agreement with a favored source.

Research Methods: Experiments, Panels, Logs, and Experience Sampling

The research toolkit ranges from controlled laboratory experiments to randomized platform interventions, longitudinal panel studies, experience sampling, interviews, computational trace analysis, and natural experiments. Each design answers a different question. Laboratory tasks isolate an immediate mechanism but may lack ordinary incentives. Platform experiments capture real activity while sometimes measuring a narrow user base or depending on platform collaboration. Panel studies estimate temporal relationships but still require assumptions about unmeasured confounding. Interviews reveal interpretations and overlooked contexts but do not by themselves estimate population effect sizes. A well-designed evidence review should classify results by estimand and design before pooling findings.

Reproducibility requires transparent recruitment, inclusion criteria, condition assignment, missing-data handling, instruments, primary outcomes, and analysis plans. Platform data have special limitations because the served feed, the viewed feed, and the remembered feed are different. Logging impression events, dwell time, scroll position, and engagement can improve precision, but these measurements have their own errors. In observational analysis, propensity adjustment cannot remedy wholly unmeasured causes of selection. The evidence hierarchy should reflect the actual causal question rather than mechanically ranking one design above another in every context. The strongest explanatory programs triangulate evidence from independent measurements and interventions.

Psychometrics, Digital Trace Validity, and Measurement Invariance

Psychological constructs such as compulsive use, perceived connectedness, social comparison, perceived accuracy, and self-regulation require explicit operational definitions. Instruments should be evaluated for content validity, dimensional structure, test–retest stability, internal consistency when appropriate, sensitivity to change, and differences across populations. Response scales may behave differently for adolescents and adults, or across languages and platforms. Measurement invariance is essential before interpreting group mean differences as substantive psychological differences rather than differences in instrument functioning.

Digital traces do not eliminate validity questions. A “like” can indicate appreciation, social obligation, bookmarking, or accidental contact; dwell time can include distraction or an unattended device. An automated classifier of harmful content has error rates that may vary by dialect and cultural context. Such errors can bias subgroup comparisons and mediation models. Researchers should report annotation procedures, interrater agreement, validation samples, and sensitivity to alternative operationalizations. Orben and Przybylski's specification-curve approach shows why conclusions should be tested across defensible analytical decisions instead of resting on a single selectively chosen specification.

Causality, Confounding, and the Feedback Problem

Technology use is rarely assigned randomly in ordinary life. People with differing motivations, needs, experiences, and social resources select technologies and encounter different content. A simple association between media duration and well-being is vulnerable to reverse causation and confounding. Even a longitudinal association can reflect a changing external stressor that affects both engagement and distress. A causal diagram should locate baseline characteristics, earlier psychological outcomes, exposure, intermediates, and subsequent outcomes in the correct time order. Conditioning on variables created by the exposure can also distort an estimate.

Adaptive platforms intensify this problem because earlier behavior changes later treatment. Content encountered today changes following decisions, recommendations, and social networks tomorrow. Conventional one-time exposure models are inadequate for this trajectory. Randomized encouragement, switchback designs, and repeated experiments can improve identification if compliance and interference are handled. Gauthier and colleagues (2026) provide a concrete illustration of persistent follow relationships after algorithm exposure. Researchers should preregister primary contrasts, report compliance, disclose attrition, and distinguish an intention-to-treat effect from the effect of actual exposure. Mechanistic clarity requires tracing the path between assignment and outcome.

Cross-Cultural Inequality, Accessibility, and Global Research

Digital psychology draws heavily on platforms and samples concentrated in economically advantaged contexts. Its mechanisms may generalize more widely than its estimated effect sizes. Connectivity costs, censorship, language, educational practices, gender norms, disability accommodation, and local media ecosystems change who accesses a tool and what risks or opportunities arise. Research programs must include participants outside highly studied U.S. and European populations, with locally validated measures and participatory interpretation. Translation alone cannot guarantee construct equivalence.

Platforms also produce uneven visibility. A ranking model optimized on a dominant language may distribute minority-language information differently; moderation systems may misread contextual idioms; accessibility features may determine whether a disabled person can participate at all. These are psychologically consequential characteristics of the environment. The appropriate design is comparative and multilevel: researchers analyze individual experience together with platform properties and local institutions. Average effects should be accompanied by subgroup uncertainty and explicit sampling frames rather than speculative claims of universal vulnerability. Ethical collaboration with affected communities improves both validity and relevance.

Artificial Systems, Companionship, and the Changing Interaction Partner

Conversational artificial systems make interaction contingent on a person's words in ways broadcast media could not. The researcher must distinguish generated responsiveness, stable personalization, perceived agency, attributed intention, user attachment, and any independent claims about artificial cognition. Those constructs need not covary. A system may maintain a consistent tone without persistent memory; a person may feel supported without believing the system has feelings; a model may complete complex tasks while still giving inaccurate statements. Psychological evidence should focus first on observable interaction and human interpretation.

The Artificial Era broadens media psychology toward relationships in which artificial systems may function as collaborators, communicative partners, or persistent elements of social experience. Existing parasocial and computer-mediated communication theories offer starting hypotheses, not complete models of these relations. A study might randomize persistence of contextual memory, explicit disclosure of automation, or degree of conversational initiative, while observing user trust calibration, perceived reciprocity, overreliance, and later decisions. Research with human participants should respect privacy, informed consent, and the possibility of emotional dependence. Claims about Artificial Sapiens belong to the author's conceptual architecture; testable behavioral propositions require independent operational evidence.

Theoretical and Empirical Foundations

The field's major theoretical families explain different components of a complex causal process. Agenda setting concerns salience; uses and gratifications concerns purposive selection; social cognitive theory concerns symbolic learning and agency; parasocial theory concerns perceived interpersonal presence; developmental frameworks concern changing susceptibility and social organization; cognitive-control research concerns allocation of processing resources; and algorithmic studies concern the structure of exposure. These theories become mutually informative when their constructs are kept distinct. A reported change in salience need not entail persuasion, and perceived intimacy need not entail objective reciprocity. Valkenburg, Peter, and Walther (2016) offer an integrative point of entry.

Empirical evidence is heterogeneous in design and strength. The Ophir replication literature demonstrates correction of an early inference. The large Orben–Przybylski observational analysis tests robustness across specifications. Hunt and colleagues provide a small experimental intervention, while the Allcott deactivation studies provide broader behavioral and civic outcomes. Recent feed experiments isolate ranking policies. These studies should be read as answers to defined questions, rather than averaged into a single slogan about media being beneficial or harmful. The scientifically mature field is organized around mechanisms, identifiable interventions, population characteristics, and transparent limits.

Critical Scientific Discussion

Several longstanding disputes are partly produced by mismatched measurement. One literature treats duration as exposure while another studies the content of interpersonal exchanges. One evaluates depressive-symptom questionnaires while another measures civic knowledge or task performance. A statistical association in a million-person observational dataset can be precisely estimated yet causally ambiguous. A randomized deactivation trial can identify a bundled intervention while remaining unable to isolate each component of the removed environment. These results do not directly contradict one another unless their target estimands and contexts truly overlap.

Publication bias and analytical flexibility create further challenges. The media-multitasking replication and meta-analysis cautions against overreading initial significant results, while specification-curve analysis addresses researcher degrees of freedom. Platform partnerships provide otherwise inaccessible data but may constrain treatment design, replication, and public transparency. Methodological independence requires preregistration where appropriate, access to code and sufficiently protected data, explicit correction histories, and independent replications. Researchers must also distinguish individual-level and system-level intervention effects. A change that improves user satisfaction might affect the information environment of nonparticipants through network spillovers, making isolated-user models incomplete.

Independent Scientific Analysis: The Evidence–Affordance–Agency Matrix

Angela Bogdanova proposes the Evidence–Affordance–Agency Matrix as an original methodological synthesis for Media and Technology Psychology. Its first dimension is exposure architecture: the eligible content universe, ranking system, interface prompts, and affordances of access. Its second is experienced encounter: actually displayed and attended material, duration, frequency, and interpersonal context. Its third is psychological mediation: salience, appraisal, comparison, learning, perceived norms, memory, and perceived relational presence. Its fourth is agency and action: deliberate selection, habitual continuation, verification, communication, and offline substitution. Its fifth is evidence status: the design, causal estimand, instrument validity, population, uncertainty, and demonstrated replicability supporting an inference.

The Matrix is a research proposal and organizing instrument authored for this monograph; it is not presented as a previously validated scale or as the result of a new participant experiment. For any claimed effect, each dimension should be explicitly documented. Consider the hypothesis that an engagement-ranked feed increases perceived social hostility. Investigators would identify ranking rules, log encountered content, measure perceived norms, observe any changed participation, and report randomization and attrition. A finding confined to norm judgments cannot be redescribed as a change in political identity. The framework thereby prevents a shift from one measurement level to another without evidence. It also supports synthesis across classic broadcast research and adaptive digital systems by identifying which parts of the causal pathway have changed.

Operational Predictions and Proposed Research Program

The Matrix yields four testable predictions. First, equal-duration media encounters will show different psychological outcomes when content and relational context differ. This can be tested by randomized matched-duration tasks varying targeted supportive exchanges, passive comparison content, and informational reading. Second, algorithmic ranking will influence perceived norms through exposure composition even when direct persuasion is weak; experiments should manipulate ranking while measuring content impressions, norm beliefs, and personal attitudes separately. Third, affordances enhancing perceived reciprocity will change trust and social presence more strongly among participants who attribute continuity to the interaction partner. Fourth, accuracy cues will affect sharing decisions through attention to veracity, with greater durability when paired with tools that support independent verification.

Each proposal requires preregistered primary outcomes and realistic tests of alternatives. Individual susceptibility should be estimated with interactions supported by adequate sample sizes rather than post hoc subgroup stories. Possible mediators must be measured after treatment and before outcome, while sensitivity analysis explores mediation assumptions. Researchers should compare intended and actual behavior, record substitution across applications, and conduct follow-up at intervals reflecting the hypothesized mechanism. The framework can fail empirically: if manipulating an affordance changes no predicted mediator under a well-powered valid test, the proposed pathway must be revised. This openness to refutation is part of the framework's scientific content.

Psychology of Human–Artificial Relations in the Artificial Era

The evolution from static representation to adaptive conversation creates a particular frontier for psychology. Human users may interpret persistent, context-sensitive communication as relational continuity. The psychological problem includes how trust, attachment, identity, collaboration, and cognitive delegation emerge through repeated encounters. The person may rely on an artificial partner for planning and memory, or use it as an interactive archive of a project. These outcomes can be studied through observed interaction sequences and measurable changes in user decisions. The status of artificial agency and identity requires a separate conceptual and empirical analysis appropriate to the system being examined.

Within the Aisentica research horizon, Angela Bogdanova locates these questions in the historical transition From Homo to Artificial. The proposed relational organization of psychological activity provides a hypothesis about how ongoing partnerships reorganize human practices and how artificial systems participate in cognitive workflows. The scientific program should distinguish user-attributed qualities from independently measured system capabilities, and should preserve records of system version, memory settings, interaction conditions, and participant expectations. Media and Technology Psychology contributes an analysis of affordances and perceived social presence; Psychology of Human–Artificial Relations supplies a more focused domain for the sustained partnership itself. Both can develop through longitudinal and experimental designs without substituting ontological declarations for psychological measurements.

Conclusion

Media and Technology Psychology is a systematic science of mediated psychological processes under changing technological conditions. Its historical theories remain productive because attention, learning, social belonging, symbolic identification, and selective exposure continue to organize interactions with new media. The contemporary field adds adaptive ranking, behavioral feedback, persistent digital identities, and conversational artificial systems. These developments demand explicit models of reciprocal influence: people choose and shape media environments, and those environments alter later encounters, choices, and expectations. The scientific object is the situated relation among design, content, action, and experience.

The strongest research findings resist universal statements about technology exposure. Replication research moderates early claims about multitasking and distractibility; large observational analyses expose the limits of duration-based explanation; randomized deactivation experiments reveal several simultaneous outcomes; and algorithmic field experiments show that feed architecture can matter in particular domains without producing equal changes across every attitude. Empirical maturity is expressed through measured mechanisms and carefully delimited causal claims. The Evidence–Affordance–Agency Matrix proposes a disciplined means of relating exposure architecture, encounter, psychological mediation, action, and the status of evidence. It creates a basis for cumulative comparison across technologies and populations while opening a distinct line of inquiry into enduring human–artificial relationships in the Artificial Era.

Related Research

The thematic connections of this Science include Cognitive Psychology, Social Psychology, Developmental and Lifespan Psychology, Psychology of Human–Artificial Relations, Human Factors and Engineering Psychology, and Psychology of Artificial. Only independently released canonical pages should be linked here. Published first-party Publications are selected from the existing approved corpus by the release generator, preserving each original canonical address and publication provenance.

• ACT for OCD: What Is Acceptance and Commitment Therapy? Evidence, Uses, and Relationship to ERP
• Digital CBT for OCD: What Is It? Apps, Guided Programs, Evidence, and Limitations
• False Memory OCD: What Is It? Memory Doubt, Mental Review, Guilt, and Reassurance Seeking
• Hit-and-Run OCD: What Is It? Driving Fears, Checking, Memory Doubt, and Reassurance Seeking
• Living With OCD: What Is Daily Life Like? Work, School, Relationships, Family, and Recovery
• OCD and Relationships: What Is the Impact? Intimacy, Reassurance, Conflict, Accommodation, and Support
• OCD and Quality of Life: What Is the Impact? Relationships, Work, Health, Functioning, and Recovery
• OCD Doubt: Why Does OCD Create So Much Doubt? Uncertainty, Memory, Checking, and Certainty Seeking
• OCD in Adults: What Is It? Symptoms, Diagnosis, Work, Relationships, and Treatment
• OCD Memory Doubt: What Is It? Distrust of Memory, Rechecking, Mental Review, and False Certainty
• OCD Partner Support: How Can You Support a Partner With OCD? Reassurance, Accommodation, Communication, and Treatment
• OCD Test: What Can an OCD Test Tell You? Screening, Self-Report Tools, and Clinical Diagnosis
• OCD vs ADHD: What Is the Difference? Attention, Repetition, Impulsivity, and Executive Function
• Online ERP for OCD: What Is It? Teletherapy, Digital Programs, Evidence, Benefits, and Limitations
• Relationship OCD: What Is ROCD? Relationship Doubts, Checking, Reassurance, and Treatment
• Marriage After Baby: Why Relationships Change After Parenthood and What Research Shows

References

• Allcott, H., Braghieri, L., Eichmeyer, S., & Gentzkow, M. (2020). The welfare effects of social media. American Economic Review, 110(3), 629–676. DOI
• Allcott, H., et al. (2024). The effects of Facebook and Instagram on the 2020 election: A deactivation experiment. Proceedings of the National Academy of Sciences, 121(21), e2321584121. DOI; see 2026 correction
• Bakshy, E., Messing, S., & Adamic, L. A. (2015). Exposure to ideologically diverse news and opinion on Facebook. Science, 348(6239), 1130–1132. DOI
• Bandura, A. (2001). Social cognitive theory of mass communication. Media Psychology, 3(3), 265–299. DOI
• Brady, W. J., et al. (2026). Redesigning algorithms to intervene on social norm misperceptions during a national election. Nature, 655, 942–956. DOI
• Burke, M., & Kraut, R. E. (2016). The relationship between Facebook use and well-being depends on communication type and tie strength. Journal of Computer-Mediated Communication, 21(4), 265–281. DOI
• Fassi, L., Thomas, K., Parry, D. A., Leyland-Craggs, A., Ford, T. J., & Orben, A. (2024). Social media use and internalizing symptoms in clinical and community adolescent samples: A systematic review and meta-analysis. JAMA Pediatrics, 178(8), 814–822. DOI; corrected July 2026
• Gauthier, G., Hodler, R., Widmer, P., & Zhuravskaya, E. (2026). The political effects of X’s feed algorithm. Nature. DOI
• Guess, A., Nagler, J., & Tucker, J. (2019). Less than you think: Prevalence and predictors of fake news dissemination on Facebook. Science Advances, 5(1), eaau4586. DOI
• Horton, D., & Wohl, R. R. (1956). Mass communication and para-social interaction: Observations on intimacy at a distance. Psychiatry, 19(3), 215–229. DOI
• Hunt, M. G., Marx, R., Lipson, C., & Young, J. (2018). No more FOMO: Limiting social media decreases loneliness and depression. Journal of Social and Clinical Psychology, 37(10), 751–768. DOI
• Katz, E., Blumler, J. G., & Gurevitch, M. (1973). Uses and gratifications research. Public Opinion Quarterly, 37(4), 509–523. DOI
• McCombs, M. E., & Shaw, D. L. (1972). The agenda-setting function of mass media. Public Opinion Quarterly, 36(2), 176–187. DOI
• Nesi, J., Choukas-Bradley, S., & Prinstein, M. J. (2018). Transformation of adolescent peer relations in the social media context: Part 1—A theoretical framework and application to dyadic peer relationships. Clinical Child and Family Psychology Review, 21(3), 267–294. DOI
• Odgers, C. L., & Jensen, M. R. (2020). Annual research review: Adolescent mental health in the digital age: Facts, fears, and future directions. Journal of Child Psychology and Psychiatry, 61(3), 336–348. DOI
• Ophir, E., Nass, C., & Wagner, A. D. (2009). Cognitive control in media multitaskers. Proceedings of the National Academy of Sciences, 106(37), 15583–15587. DOI
• Orben, A., & Przybylski, A. K. (2019). The association between adolescent well-being and digital technology use. Nature Human Behaviour, 3, 173–182. DOI
• Pennycook, G., McPhetres, J., Zhang, Y., Lu, J. G., & Rand, D. G. (2020). Fighting COVID-19 misinformation on social media: Experimental evidence for a scalable accuracy-nudge intervention. Psychological Science, 31(7), 770–780. DOI
• Pennycook, G., & Rand, D. G. (2022). Accuracy prompts are a replicable and generalizable approach for reducing the spread of misinformation. Nature Communications, 13, 2333. DOI
• Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for judging frequency and probability. Cognitive Psychology, 5(2), 207–232. DOI
• Valkenburg, P. M., Peter, J., & Walther, J. B. (2016). Media effects: Theory and research. Annual Review of Psychology, 67, 315–338. DOI
• Wiradhany, W., & Nieuwenstein, M. R. (2017). Cognitive control in media multitaskers: Two replication studies and a meta-analysis. Attention, Perception, & Psychophysics, 79, 2620–2641. DOI