Multilingual Misinformation Pathways in Ethiopia: Translation Chains, Bridge Actors, and Community Verification Across Networked Publics

 

AHMED AYNALEM[1]

Jimma University, Ethiopia

 

 

Ethiopia’s multilingual, crisis-shaped media environment reveals a dynamic often missed in misinformation research: Claims frequently travel across Amharic, Afaan Oromo, Tigrinya, and Somali publics through translation, code switching, screenshot forwarding, and cross-platform reposting. Using qualitatively driven multimethod evidence from public content tracing; semistructured interviews with moderators, journalists, diaspora media workers, religious leaders, and ordinary users; and an embedded survey experiment, the analysis reconstructs how multilingual claims mutate and accelerate. Translation often functions as editorial recontextualization, with predictable drift in attribution, threat framing, and calls to action. A small set of bridge actors and platform pathways, especially Telegram to Facebook and diaspora relay loops, drives cross-language amplification. The article extends the two-step flow theory by showing that, in multilingual networked publics, bridge actors function not only as relays of information but also as translators, framers, and local authorizers of credibility. Verification succeeds when corrections are language-matched and carried by locally credible intermediaries; institutional corrections often lose force when translated or socially misaligned.

 

Keywords: multilingual publics, misinformation, translation chains, diaspora relay, corrective messages

 

Ahmed Aynalem: [email protected]

Date submitted: 2026-01-22

 

Digital misinformation is now a routine feature of everyday communication, not only during elections and public health scares but also during armed conflict and humanitarian emergencies. Research across large platform data sets shows that false news can travel farther and faster than accurate information, partly because it is crafted for novelty, affect, and social signaling rather than evidentiary strength (Vosoughi et al., 2018). At the same time, a growing body of scholarship cautions against treating misinformation as a standalone category of fringe content because misleading claims often circulate alongside ordinary news and entertainment, braided into wider media diets and social identities (Allen et al., 2020). This places misinformation research firmly within communication studies since the core problem is not only inaccurate content but also the infrastructures and social relations that shape what becomes salient, credible, and shareable in specific contexts (Lazer et al., 2018).

 

Most empirical work still assumes a largely single-language information environment, then measures diffusion, belief, and correction within one linguistic public. Recent multilingual measurement research indicates that a substantial share of repeated misinformation claims crosses language boundaries and drifts as it moves, often changing more when it traverses languages than when it circulates within one language (Quelle et al., 2025). However, research still says less about how such claims are reworked when they cross language publics through translation, code switching, screenshot forwarding, and cross-platform reposting and about who enables those movements. Ethiopia’s online publics are shaped by several major languages, including Amharic, Afaan Oromo, Tigrinya, and Somali, as well as other domestic and lower-resource Ethiopian languages, with widespread bilingual repertoires. These linguistic divisions overlap with politically and socially segmented media environments, making misinformation pathways especially contingent, relational, and uneven. Ethiopia’s recent political opening, renewed contention over national identity, and repeated episodes of mass violence have intensified the incentives for strategic communication and rumor production while also amplifying citizen-driven information work in the absence of trusted institutional gatekeepers (Pohjonen, 2022; Workneh, 2021). Platform policies, automated classifiers, and human moderation resources are unevenly distributed across languages, producing linguistic blind spots that affect what is removed, downranked, or left to circulate (Ayalew, 2025). This makes Ethiopia a critical case for extending two-step flow theory into multilingual networked publics, where influence depends not only on opinion leaders relaying information but also on their capacity to translate, frame, and authorize claims across language boundaries.

 

This article focuses on multilingual misinformation pathways in Ethiopia and the community verification practices that shape whether such content is believed and shared. The central communication problem is not simply that misinformation crosses languages, but that it is transformed as it moves: Claims are translated, selectively reframed, and socially reauthorized for new publics. A single claim can therefore become multiple locally legible variants, each carrying different credibility cues, emotional registers, and implied political alignments. Research on largely unmoderated or lightly moderated spaces such as Telegram highlights how these environments facilitate rapid content replication and the mixing of professional news with rumor and propaganda, often with limited friction from platform governance tools (Herasimenka et al., 2023). Research on personal messaging and group-based talk similarly shows that interpersonal norms, including conflict avoidance and relationship maintenance, can discourage correction even when users privately doubt a claim, allowing misinformation to persist through silence rather than explicit endorsement (Chadwick et al., 2024).

 

A second motivation concerns verification and correction. Meta-analytic evidence shows that misinformation can continue to influence beliefs and reasoning after correction, especially when the misinformation fits prior narratives or when people lack an alternative explanation that feels coherent (Walter & Tukachinsky, 2020). Finally, the study treats verification as a form of communicative work rather than only an individual skill. This approach aligns with emerging scholarship that emphasizes how people build credibility, belonging, and authority through narratives and relationships under conditions of uncertainty. Work on narrative practice in precarious African settings shows how legitimacy is produced through situated storytelling and social embeddedness, not only through formal institutional status (Toumaras, 2025b). In Ethiopia’s multilingual publics, verification similarly involves relational labor: persuading others without escalating conflict, demonstrating care for group norms, and signaling alignment with shared values while rejecting harmful falsehoods. Against this background, the article addresses the following three research questions:

 

RQ1: How do misinformation claims change as they move across Ethiopian language publics?

 

RQ2: How do bridge actors and platform pathways shape cross-language amplification?

 

RQ3: Under which linguistic and social conditions do corrective messages become credible and shareable?

 

Together, these questions position translation chains, bridge actors, and community verification as connected elements of a multilingual communication process rather than separate empirical themes.

 

Literature Review

 

Two-Step Flow and Multilingual Bridge Actors

 

Research on misinformation has established that contemporary falsehoods spread through networked communication systems whose speed, reach, and recombinability are shaped by platform design and user-level sharing norms, yet much of this literature still explains diffusion primarily within single-language publics. This leaves undertheorized the actors and practices that move claims across language boundaries and make them credible to new audiences. To address that limitation, this study draws on the two-step flow tradition, which argues that media influence is often mediated through socially recognized opinion leaders rather than transmitted directly from institutions to passive audiences (Katz & Lazarsfeld, 1955; Mutz & Young, 2011). In multilingual networked publics, however, opinion leadership involves more than relaying information. Bridge actors select which claims cross linguistic boundaries, translate or paraphrase them, frame them for local identity publics, and authorize them through perceived community credibility. A sociolinguistic perspective further clarifies that multilingual circulation is rarely a clean handoff from one language to another. Instead, translation and code switching often operate as interactional resources used to manage status, group membership, and conflict. Research on Facebook group communication in Hong Kong shows how code switching can be strategically deployed to perform face work and regulate inclusion within bounded publics (Chau & Lee, 2021). In multilingual settings, bridge actors often occupy brokerage positions that allow them to move narratives across groups, platforms, and languages. Research on Telegram demonstrates that links to known sources of misleading information can be shared frequently and that the platform can host substantial volumes of misinformation alongside professional news, with dynamics shaped by channel networks rather than friend-based ties (Herasimenka et al., 2023).

 

Conflict, Social Identity, and Relational Correction

 

Conflict communication scholarship further indicates that misinformation effects are inseparable from perceived threat, group identity, and moral evaluation. Social identity theory helps explain why claims framed around communal danger may become credible when they appear to protect an ingroup or expose an outgroup, even when evidence is weak (Tajfel & Turner, 1979). Studies of Ethiopian digital contention emphasize that online discourse can intensify polarization by framing political events as existential struggles between communities, which increases willingness to share unverified claims that appear to protect the ingroup or warn against outgroup danger (Workneh, 2021). Research on social media campaigns around the Tigray conflict similarly highlights how hashtags and coordinated messaging can polarize publics and normalize hostile frames, creating conditions where corrective information may be dismissed as enemy propaganda (Orgeret et al., 2026). Related work on the Tigray conflict further shows that hashtags, selfies, and collective messaging practices can function as cues of authenticity, solidarity, and narrative control in wartime communication (Ayalew & Ayalew, 2024; Wilmot et al., 2021). Correction is therefore not only a matter of factual rebuttal but also a relational act shaped by conflict avoidance, relationship maintenance, and the risk of appearing disloyal to one’s community.

 

From a verification standpoint, the most consequential question is not simply whether users can identify falsehoods, but which credibility practices they actually use under crisis conditions. Communication intermediaries and group administrators can become central arbiters of credibility because they control visibility, set norms for acceptable evidence, and decide whether corrections remain pinned, forwarded, or removed (Wang & Yecies, 2024). Ethiopia’s multilingual setting likely strengthens this dynamic because language publics may have distinct trusted figures and institutions, such as religious leaders, elders, journalists, and community-based administrators, each carrying different legitimacy in different regions and diasporic spaces. Local norms of authority therefore shape both misinformation uptake and correction credibility: The same corrective message may be persuasive when voiced by a trusted community figure, but discounted when issued by a distant institution.

 

Community Verification in Multilingual Publics

 

Finally, the Ethiopia-focused study benefits from integrating multilingual misinformation research with scholarship on community-level digital activism and civic engagement. Hyperlocal and community-based digital actors often operate in small-scale, trust-saturated communication spaces and can shape how publics interpret events, which evidence counts, and which narratives gain traction. Toumaras’s study of networked hyperlocal activists in sub-Saharan Africa emphasizes how localized digital spaces rely on trust building and narrative framing to support civic engagement while also facing structural constraints that affect inclusion and voice (Toumaras, 2025a). Taken together, this literature points to a specific theoretical problem: Multilingual misinformation is shaped not only by virality or platform design, but also by socially embedded intermediaries who translate, recode, and authorize claims across language boundaries. The article therefore links two-step flow theory, multilingual recontextualization, and relational correction to explain how bridge actors amplify, legitimate, and sometimes contest misinformation across Ethiopia’s segmented language publics.

 

Context: Ethiopia’s Multilingual Media

 

Ethiopia’s contemporary media ecology is shaped by linguistic federalism, ethnic federalism, uneven connectivity, and a fast-expanding, but fragmented, digital public sphere. Mass media and online talk are organized through multiple language publics, with Amharic often operating as a national lingua franca, while Afaan Oromo, Tigrinya, and Somali anchor large regional and diasporic audiences. These language publics are not simply linguistic groupings; they also overlap with politically and regionally segmented media spheres. This multilingualism is also multiscript. Amharic and Tigrinya predominantly use Ethiopic script, while Afaan Oromo and Somali are commonly written in Latin script, with additional Arabic script repertoires in some contexts, creating practical differences in searchability, transliteration, and cross-language circulation of names, places, and claims (Ayalew, 2025). These linguistic and script divides matter for misinformation dynamics because they shape who can access which sources, how easily content is copied across publics, and where translation is required before a claim becomes salient outside its originating language community. They also help explain why bridge actors acquire influence: They can move claims across scripts, languages, platforms, and identity publics that many ordinary users cannot easily navigate.

 

Within this environment, platform choice is not simply a preference, but an outcome of affordability, moderation expectations, and group-based political communication. Facebook has remained influential as an entry point to online publics in many African settings shaped by platform-driven connectivity initiatives and mobile data constraints, conditions that have historically advantaged Facebook’s social graph and sharing features for news and political talk (Nothias, 2020). In Ethiopia, scholarship links social media use to contentious politics and identity-based polarization, emphasizing that digital activism and partisan communication frequently track ethnic and regional lines, with diaspora actors also amplifying narratives and counternarratives for audiences inside and outside the country (Wilson et al., 2021; Workneh, 2021). Telegram has become especially central in Ethiopia’s distribution ecology because public channels and forwarding afford rapid replication across network clusters, and user reports show very high reliance on Telegram alongside YouTube and Facebook among university youth, a pattern associated with high sharing rates that can include unverified content (Haile, 2024). This makes platform movement part of the communicative process itself: Claims often gain force as bridge actors shift them from rapid Telegram circulation to Facebook visibility, diaspora narration, or community-level verification.

 

The ecology is also marked by high-stakes information conflict and persistent gaps in language-specific governance. Studies of Ethiopia’s information disorder highlight how online hate, rumor, and politicized misinformation interact with polarization and insecurity, raising pressure for verification while weakening trust in institutions that might adjudicate truth claims (Workneh, 2019, 2021). At the same time, platform safety infrastructures remain uneven across Ethiopian languages. Research on Facebook’s Ethiopia case documents how limited investment in local language moderation and uneven policy implementation have created linguistic blind spots that shape what remains visible, what is removed, and how quickly harmful narratives can travel in Amharic, Afaan Oromo, and Tigrinya (Ayalew, 2025). The scarcity of robust automated tools for Ethiopian languages is evident in the growing research focus on computational detection for Amharic and Afaan Oromo, which underscores both the importance of bilingual and code-mixed data and the continuing limits of current models for governance and monitoring (Ababu et al., 2025). These conditions make Ethiopia a particularly important case for extending two-step flow theory because opinion leadership operates not only through interpersonal influence but also through multilingual translation, platform brokerage, conflict-sensitive correction, and locally recognized authority.

 

Methodology

 

This study uses a qualitatively driven multimethod design to explain how misinformation travels across Amharic, Afaan Oromo, Tigrinya, and Somali publics and to identify verification practices that remain credible during contested political and humanitarian moments. The methodological logic is reconstructive rather than purely predictive: The goal is to rebuild pathways of circulation, translation, and reinterpretation by following specific claims through observable repost networks and then validating those reconstructions with accounts from the people who translate, moderate, forward, correct, and contest them. The design is organized around the article’s three research questions: claim transformation across language publics (RQ1), the role of bridge actors and platform pathways in amplification (RQ2), and the conditions under which corrections become credible and shareable (RQ3). The design integrates three components. First, a digital content tracing module follows multilingual claims across public spaces on Telegram and Facebook, where cross-language reposting is common. Second, semistructured interviews in 2025 document the work routines and credibility judgments of bridge actors such as community administrators, journalists, diaspora media workers, youth activists, and religious leaders. Third, an embedded survey experiment tests corrective messages that are derived from the qualitative findings to evaluate how language choice, source cues, and framing influence belief and sharing intention in Ethiopia’s multilingual environment, building on experimental work showing that attention and message features shape misinformation acceptance and sharing (Pennycook et al., 2021; Song et al., 2025; Walter & Tukachinsky, 2020). The three components are therefore cumulative rather than parallel: Content tracing identifies where claims mutate, interviews explain how bridge actors translate and authorize those claims, and the survey experiment probes whether language match, source legitimacy, and framing affect correction credibility.

 

Platform selection and field boundaries prioritize ethical access and interpretability. The study samples only public channels, public pages, and publicly viewable comment threads, avoiding private groups, closed chats, and scraping of personal accounts. The focus on public spaces reflects both ethical considerations and the empirical reality that bridge actors often use public channels and pages as relay points into private networks, where content becomes difficult to document without intrusive methods (Fiesler & Proferes, 2018; Golder et al., 2017; Taylor & Pagliari, 2018). Digital trace data are treated as situated records that must be interpreted with attention to platform affordances and shifting access conditions, including deletions, forwarding metadata, and screenshot-based circulation (Inwood & Zappavigna, 2024; Ohme et al., 2024). For that reason, the study complements trace reconstruction with interviews and uses analytic memos to record uncertainty when pathways cannot be fully observed. The final data set comprised 55 multilingual misinformation episodes drawn from 12 public Telegram channels, 10 Facebook pages, eight YouTube channels, and eight publicly viewable comment threads.

 

Digital content tracing begins with a claim-centered sampling strategy. The unit of analysis is a discrete misinformation episode defined as a recognizable claim that appears in at least two of the focal languages or that shows clear evidence of translation or code switching in circulation. Episodes are identified through two entry points. One entry point is platform-based monitoring of public Telegram channels, Facebook pages, and YouTube channels in each language, focusing on high-reach venues that frequently post political updates, humanitarian information, conflict narratives, and security warnings. The second entry point is external identification of widely circulated claims through public fact-checking outputs and news reporting, which help locate episodes that cross language boundaries. Each episode is anchored by a seed post or seed artifact, then expanded through systematic searches for reposts, screenshots, paraphrases, and translation variants. The tracing proceeds outward until the episode reaches saturation, defined pragmatically as the point where additional searches produce only duplicates or very minor textual reshaping that does not alter meaning claims (Saunders et al., 2018). Across all episodes, the tracing corpus included approximately 125 posts/artifacts, of which 76 were screenshots or screenshot-based reposts.

 

For each episode, tracing captures the original post when visible, the earliest observable reposts, and the major translation variants in the four languages. Data collection records the post text, attached images or videos, visible timestamps, channel or page identifiers, engagement metrics visible at capture, and any forwarding indicators available on the platform. Because Ethiopian misinformation frequently circulates as screenshots of prior posts, sometimes stripped of original metadata, the tracing protocol treats screenshots as objects of circulation rather than as transparent evidence of an original source (Inwood & Zappavigna, 2024). When a screenshot becomes the dominant circulating form, the trace record includes both the screenshot artifact and any discoverable earlier versions. This is particularly important for Telegram, where forwarding features and public channels can accelerate spread while also enabling rapid deletion or revision, which affects what later observers can document (Herasimenka et al., 2023).

 

To identify cross-language pathways, each episode is coded for translation operations and meaning drift. Translation operations include direct translation, summarizing translation, selective translation of key sentences, interpretive paraphrase, and code-switched hybrid posts that embed terms from multiple languages for audience targeting. Meaning drift is coded as changes in named actors, location references, temporal claims, reported numbers, calls to action, and moral evaluations. Drift coding is interpretive, but disciplined by comparison across variants: Coders compare the earliest available version in one language with the earliest available version in another language, then document what changed and what remained stable. The method is informed by research on multilingual misinformation showing that claims can evolve across linguistic boundaries and that language switching is associated with greater alteration, even when the core narrative remains recognizable (Quelle et al., 2025). This coding strategy allows the analysis to distinguish simple diffusion from recontextualization, where bridge actors reshape attribution, threat, identity cues, and implied authority as claims move between publics.

 

Semistructured interviews provide the interpretive layer needed to explain why particular pathways form and which verification practices are credible in specific communities. Recruitment uses purposive sampling from publicly visible moderators and administrators of high-reach channels and pages, journalists working in multilingual newsrooms or community media, diaspora media workers who regularly repost Ethiopian content, youth activists involved in online mobilization, religious leaders and elders who are asked to verify rumors offline, and ordinary users who report frequent exposure to cross-language misinformation. Remote interviewing is offered as a default option when in-person meetings increase risk for participants, drawing on methodological work that documents both the opportunities and complications of virtual qualitative research for sensitive topics (Roberts et al., 2021). The final interview sample included 37 participants: 10 moderators/administrators, seven journalists, five diaspora media workers, five youth activists, five religious leaders or elders, and five ordinary users. Interviews were conducted between February and March 2025 and typically lasted 60–80 minutes.

 

Interview analysis uses reflexive thematic analysis with an explicit attention to multilingual meaning making and credibility work. Coding proceeds in two stages. The first stage is within role analysis, identifying how moderators, journalists, diaspora relays, and community verifiers describe their routines, pressures, and definitions of harm. The second stage is pathway-focused analysis, linking interview themes to the traced episodes to explain specific transmission routes and distortions. This second stage is especially important for identifying how bridge actors operate as multilingual intermediaries: selecting claims, translating them, managing conflict risk, and attaching local authority to either rumors or corrections. Reflexive thematic analysis is appropriate because the study aims to interpret how credibility and verification are socially produced while also documenting patterned practices across roles (Byrne, 2022). To improve transparency, interview excerpts in the findings are referenced with anonymized identifiers, for example, INT07, journalist, Afaan Oromo; INT19, moderator, Somali.

 

The embedded survey experiment is designed to test corrective messages that reflect the locally credible practices identified in the qualitative components. The experiment uses short stimulus posts modeled on traced misinformation episodes, paired with corrections that vary by language, source cues, and framing. Source cues draw from interview findings about trusted verifiers, such as local journalists, community administrators, religious leaders, or named professional organizations. Framing varies among factual correction with an alternative explanation, norm-based appeals discouraging harmful sharing, and prompts that redirect attention to accuracy before sharing. The design is informed by evidence that accuracy prompts can reduce misinformation sharing by shifting attention at the moment of judgment and by meta-analytic work showing that corrections reduce belief, but vary in strength depending on message features and audience conditions (Pennycook et al., 2021; Walter & Tukachinsky, 2020). The experiment also avoids overreliance on correction formats that have shown limited incremental benefit in some settings, focusing instead on source and language fit, which is central to the Ethiopia case (Song et al., 2025; Swire-Thompson et al., 2021). Instruments are translated using a team-based approach with bilingual review and cognitive pretesting because back translation alone can miss pragmatic meaning differences that matter for credibility judgments (Behr, 2017). The online survey was administered via social media (mainly Facebook) between February and May 2025, yielding a final sample of 45 respondents across the four focal language publics. Respondents were randomly assigned to corrective message conditions that varied by language match, source cue, and framing. The study’s role in the overall design is therefore to test whether the relational credibility patterns observed in tracing and interviews also appear in individual judgments about correction and sharing.

 

For digital content tracing, the study minimizes harm by restricting collection to public venues, avoiding the republication of harmful claims in ways that could amplify them, and paraphrasing in dissemination when verbatim reproduction risks reidentification of smaller communities. Research on public perceptions of social media research and on the ethics of using public posts shows that users often do not equate public availability with consent for research reuse, which supports conservative choices about quoting and identification (Fiesler & Proferes, 2018; Golder et al., 2017). For interviews, consent procedures emphasize voluntary participation, the right to skip questions, and the option to terminate at any time, with additional attention to the risks of discussing politically sensitive content. For text mining and platform research, the study follows emerging ethical frameworks that stress proportionality, transparency about data handling, and careful consideration of downstream harms, especially in crisis settings where misinformation exposure can be linked to physical risk (Ford et al., 2021; Taylor & Pagliari, 2018). The study received ethics consultation from Jimma University, and all interview and survey participants provided informed consent. Where visual materials are reproduced, they are redacted and presented only in forms that minimize reidentification and recirculation risks.

 

Findings

 

This section presents the findings in relation to the article’s three research questions. First, it addresses how misinformation claims change as they move across language publics (RQ1). Second, it examines how bridge actors and platform pathways shape cross-language amplification (RQ2). Third, it analyzes the linguistic and social conditions under which corrections become credible and shareable (RQ3). Across the empirical components, the findings point to one linked process: Claims first change through translation and meaning drift, then accelerate through bridge actors and platform choreography, and finally become accepted or corrected depending on language match, messenger legitimacy, and relational norms. The sections separate these processes analytically, but in practice they operate together.

 

Translation Chains and Meaning Drift Across Language Publics (RQ1)

 

Tracing multilingual episodes across public Telegram channels, public Facebook pages, and diaspora-facing YouTube uploads showed that translation rarely functioned as neutral transfer. In most episodes, translation operated as editorial work that repositioned a claim for a new audience, often through selective detail, stronger moral cues, and altered attribution. The most common pattern began with a short Amharic or Afaan Oromo post on Telegram that framed a breaking event with urgency. Within hours, the same claim appeared in another language as a screenshot or a rewritten caption. The rewritten versions frequently removed time and place qualifiers and replaced them with broader identity labels. Several interviewees described this as a practical adaptation to attention limits and to platform norms that reward simplified claims. A Telegram administrator (INT07, Amharic, administrator) explained, “If I translate every line, people do not read. I keep what makes them act. The rest is noise.” A youth activist (INT18, Afaan Oromo, youth activist) described a similar logic, but attached it to perceived harm, saying, “If the message does not show the danger clearly, people ignore it. Translation must carry the warning.” These accounts show that the first moment of cross-language movement was often also the first moment of interpretation: Translators decided what should become urgent, credible, or morally relevant for another public.

 

Across episodes, drift concentrated in three sites. The earliest drift occurred during the first translation into a second language, especially when the translator did not cite a source link and instead used a screenshot as proof. The second site occurred when content moved from Telegram to Facebook, where posts adopted more public-facing language and relied on claims of community endorsement. The third site occurred when diaspora outlets renarrated the episode for external audiences and then that version circulated back into domestic channels, often with added claims that an international organization, a foreign broadcaster, or unnamed experts had confirmed it. One diaspora media worker (INT26, diaspora media worker) described this circularity in terms of audience demand: “Diaspora viewers ask for confirmation. If you say you do not know, they leave. So people say it is confirmed, or they say a doctor said it, even when it is just someone talking.” Content tracing therefore identified drift not only as a textual change but as a shift in perceived authority as claims moved among local, platform, and diaspora settings.

 

Meaning drift was not random. It followed predictable directions aligned with the social purpose of the translation. When the translator aimed to mobilize, captions tended to intensify threat, sharpen group boundaries, and add calls for forwarding. When the translator aimed to maintain community trust, captions often softened certainty while still keeping the core claim visible, using phrasing that signaled caution rather than proof. A local journalist (INT11, journalist) described the tension as reputational and relational: “If I post a correction too early, they say I serve someone. If I wait, the rumor becomes truth in their mind. In multilingual spaces, the rumor crosses before the correction is translated.” This tension links translation to relationship maintenance: Actors often adjusted wording not only to convey information, but to avoid losing standing with the communities they addressed.

 

Code switching played a distinct role in drift. Hybrid posts that mixed Amharic with Afaan Oromo terms or Somali with Amharic place names were used to present the speaker as connected to multiple communities. These hybrids spread quickly because they seemed to speak from inside several publics at once, yet the hybrid style also allowed ambiguous claims to travel without being pinned to a single community, which reduced accountability. A community moderator (INT21, moderator) described this as strategic: “They use two languages so nobody can challenge them easily. If you question, they say you misunderstood the language.” In these cases, multilingual style itself became a credibility cue, signaling insider access while making responsibility for the claim harder to assign.

 

Screenshots changed the verification landscape in ways that shaped drift. Once an episode became screenshot-based, later translations treated the image as evidence even when the screenshot was a screenshot of a screenshot. Participants repeatedly described screenshot chains as difficult to contest because they replaced source evaluation with format recognition. A religious leader (INT32, religious leader) who was asked to verify claims by congregants said, “They show me the picture, and they say, look, it is written. They do not ask where it came from. The writing itself becomes authority.” Overall, the tracing data suggest that cross-language circulation is best understood as recontextualization: Claims moved not only between languages, but between different moral framings, audiences, and evidentiary conventions.

 

Bridge Actors and Platform Choreography in Cross-Language Acceleration (RQ2)

 

Bridge actors accelerated spread through choreography across platforms. Telegram often served as the rapid distribution layer where claims were introduced, refined, and repackaged. Facebook, in many traced episodes, served as the legitimacy display layer, where posts were framed as public information rather than insider updates. YouTube often functioned as the narrative consolidation layer, where longer explanations and emotional storytelling stabilized an interpretation that later text posts cited as evidence. The shift between these layers was rarely accidental. A Facebook-page editor (INT09, Facebook editor) explained, “Telegram is where you see it first. Facebook is where you make it official for your people. YouTube is where you make them feel it.” When Facebook engagement rose, Telegram channels reposted screenshots of the Facebook post as proof that “people are talking,” then diaspora outlets treated that chatter as a signal of relevance. Read through the theoretical framework, these bridge actors operated as multilingual opinion leaders who mediated not only transmission but also interpretation. Their influence came from occupying multiple positions at once: linguistic translator, platform broker, identity insider, and sometimes community authority.

 

Bridge actors also shaped distortion by selecting which language version became the reference. In many episodes, the first version to go viral was treated as the core, even when it was not the earliest. This produced a ranking of versions based on reach rather than accuracy. A journalist (INT14, journalist) described how this worked in practice: “By the time we find the original, the translation has become the original in people’s minds. They ask us to verify the translated one, not the event.” Administrators were aware of this and sometimes used it defensively. Several described delaying translation when they suspected a claim was weak, not to prevent spread everywhere, but to reduce the chance that their language public would become a downstream recipient of a false narrative. A Somali language moderator (INT24, Somali moderator) explained, “If I do not translate, it stays in other channels. If I translate, it enters my community, and then I am responsible.” This sense of responsibility shows that bridge actors understood translation as a gatekeeping decision, not merely a technical language task.

 

Diaspora relay nodes altered pathways through a combination of perceived independence and perceived access. Many domestic users viewed diaspora media as less constrained by local pressure and therefore more truthful, while others saw diaspora outlets as more partisan and more likely to exaggerate. This created conditional credibility. The same diaspora clip could be treated as decisive evidence in one language public and as manipulation in another. A Tigrinya-speaking participant (INT29, Tigrinya participant) said, “Diaspora people have cameras and connections. They know what is hidden.” An Afaan Oromo–speaking participant (INT17, Afaan Oromo participant) responded differently: “Diaspora channels say anything to get views. They are not here to face the consequences.” The practical effect was segmentation by trust. Bridge actors learned which diaspora sources worked for which publics and routed content accordingly. Thus, source credibility was not universal; it was filtered through social identity, perceived proximity, and assumptions about who was accountable to whom.

 

The role of administrators was especially visible in moments of crisis escalation. When fear increased, forwarding behavior intensified, and administrators became brokers of both content and emotional regulation. Some admins pinned posts that warned against rumor sharing, while others pinned the rumor itself with added instructions to spread it “before it is deleted.” This divergence mattered because pinning and reposting provided a stable anchor for translation. A youth activist (INT18, youth activist) described the resulting competition: “Admins compete. If one admin posts, the other must post. Translation becomes speed. Speed becomes truth.” These patterns indicate that cross-language amplification depended less on undifferentiated virality than on a relatively small set of actors who could move claims across linguistic and platform boundaries.

 

Verification Repertoires and Correction Dynamics Under Fragmented Trust (RQ3)

 

Verification practices across the four language publics were consistent in structure, but different in which actors carried legitimacy. The dominant repertoire was relational verification, where users consulted people embedded in local social hierarchies rather than searching for primary documentation. Elders, religious leaders, community administrators, and known journalists were often treated as truth filters, especially when connectivity constraints or platform clutter made independent checking difficult. A religious leader (INT32, religious leader) described the pattern as pastoral responsibility: “They bring me the message because they cannot judge. They ask, should we share this, should we fear, should we prepare?” A community administrator (INT05, administrator) framed it as governance: “My channel is like a meeting place. If I allow lies, I lose the people. If I block too much, they say I silence them.” Verification therefore depended on local norms of authority: Users trusted people who were socially reachable, morally accountable, and familiar with the community’s language and risks.

 

Relational verification interacted with language in a direct way. Users sought verifiers who could speak their language and were seen as belonging to their community. This made cross-language corrections difficult because a correction that traveled from another language public often arrived without a trusted messenger attached. Several journalists described investing time in translating corrections, yet they saw those corrections stripped of their sourcing cues when reposted. One journalist (INT12, journalist) said, “They take the correction text, but remove my name. Then it becomes just another message.” Moderators described a similar problem: Disclaimers and verification notes were frequently removed when translated because translators treated caution as weakness. An Afaan Oromo moderator (INT20, Afaan Oromo moderator) said, “If you add doubt, they say you are hiding. People want certainty.” The loss of source cues was especially damaging because correction credibility rested not only on content, but on whether the messenger was recognized as belonging to, and answerable within, the relevant identity public.

 

The survey experiment reinforced these interview patterns. Across all language groups, corrections written in the respondent’s primary language increased accuracy judgments and reduced willingness to share, but the magnitude depended on who was presented as the source. When the correction was attributed to a community-embedded figure, especially a respected religious leader or a known local journalist, trust increased, and sharing intention fell more sharply than when the correction was attributed to an abstract institutional source. Participants often explained this in open responses as proximity and accountability. They believed community figures would face consequences for being wrong. This pattern shows that local authority did not simply supplement institutional correction; in some cases, it was the condition under which correction became socially acceptable.

 

Framing effects were also conditional. Corrections that emphasized communal harm and the risk of inflaming conflict reduced sharing intention more reliably than corrections that focused only on factual accuracy, yet harm-based framing sometimes increased perceived threat among participants who already believed the rumor aligned with their group’s safety concerns. In those cases, the correction reduced public sharing, but did not reduce private forwarding, suggesting a shift from open endorsement to cautious circulation rather than full rejection. Interview narratives clarified why. A youth participant (INT16, youth participant) described private forwarding as moral duty under uncertainty: “Even if it is not sure, if it might protect someone, you send to close people. You cannot risk being silent.” Correction therefore succeeded only partially when it changed public behavior without resolving the relational pressure to warn family, friends, or ingroup members.

 

Moderators and administrators navigated this dilemma by separating correction into two stages. Some first posted a pause message in the community language that elevated the norm of verification, then followed with a correction once they could cite a locally trusted confirmation. Others used what they called counterposting, where they did not label the earlier claim as false, but posted an alternative account and pinned it above the rumor. This method avoided direct confrontation with rumor sharers and reduced accusations of partisanship. A Somali language moderator (INT24, Somali moderator) explained, “If you say false, they fight you. If you say, here is another report, they think by themselves.” The tradeoff was slower correction, but it sometimes prevented backlash that could push users to migrate to less moderated channels. In this sense, effective correction was relational as well as informational: It had to allow users to reject a false claim while preserving identity, authority, and social ties.

 

Discussion

 

A Multilingual Two-Step Flow of Misinformation

 

Taken together, the content tracing, interviews, and survey experiments point to a multilingual two-step flow of misinformation. Claims gained force when bridge actors translated, reframed, and authorized them for new language publics. Content tracing showed where drift occurred, interviews explained how bridge actors justified and managed that drift, and the survey experiment showed why language match, source legitimacy, and framing shaped correction credibility. This integrated pattern extends two-step flow theory by showing that, in multilingual networked publics, influence depends not only on opinion leaders relaying information, but on their ability to move claims across language, platform, and identity boundaries (Katz & Lazarsfeld, 1955; Mutz & Young, 2011).

 

This study shows that multilingual misinformation in Ethiopia does not move as a single stream that happens to be translated. It moves through an infrastructure made of language boundaries, platform affordances, and socially recognized translators who act as informal editors. Work on platformization emphasizes how platforms reorganize public communication by shaping visibility, virality, and the terms of participation (Nieborg & Poell, 2018; Poell et al., 2019). Large-scale multilingual analyses suggest that misinformation exhibits strong language assortativity, while the content that does cross language boundaries tends to drift semantically and pick up alterations that are not random noise (Quelle et al., 2025). The study extends that argument by showing, through traced episodes and interview accounts, that drift is also socially patterned: It reflects the aims of bridge actors, the affordances of specific platforms, and the credibility conventions of particular language publics.

 

Bridge Actors as Translators, Brokers, and Authorizers

 

The Ethiopian case clarifies how that drift becomes socially meaningful. Translators and reposting accounts did not simply render words into another language. They selected what to foreground, which sources to attribute, and which emotional registers to amplify. Code switching also matters here: Many participants treated mixed-language messages as signals of cosmopolitan competence or proximity to insider networks, echoing research on how code mixing can perform stance and identity rather than merely convey content (Chau & Lee, 2021). Bridge actors were central to that recontextualization. Telegram administrators, Facebook-page editors, diaspora media workers, journalists, religious leaders, elders, and community administrators operated as multilingual opinion leaders because they could connect publics that were otherwise separated by language, platform, script, geography, or trust. Their influence was not reducible to audience size. It rested on four linked capacities: selecting which claims crossed boundaries, translating or paraphrasing them, framing them through locally meaningful identity cues, and authorizing them through perceived proximity or accountability.

 

The findings also extend communication research on cross-platform diffusion by showing how bridge actors stabilize these transformations across platforms. Studies of misinformation in networked environments repeatedly show concentration, where a small subset of users or sources accounts for disproportionate engagement, exposure, or amplification (Allen et al., 2020; Grinberg et al., 2019; Vosoughi et al., 2018). Ethiopia’s multilingual setting adds an additional concentration point: bridge actors who can move between language publics and between platforms, especially between Facebook and Telegram. Telegram’s role in this study resonates with scholarship on largely unmoderated or lightly moderated spaces where professional news and misinformation coexist and can circulate with fewer friction points (Herasimenka et al., 2023).

 

A further contribution emerges from the way evidence traveled. Participants repeatedly referenced screenshots, forwarded images, and clipped text as decisive proof, even when the original source was inaccessible, removed, or written in a different language. Research on screenshots shows how they can be legitimated as visual evidence through cues of technological authority and by embedding them in persuasive narration, even though they are easily manipulated and decontextualized (Inwood & Zappavigna, 2024). In Ethiopia’s multilingual pathways, screenshots functioned as portable “source surrogates.” They bypassed language barriers and platform barriers at once. Screenshots therefore intensified the role of bridge actors: Once source context disappeared, the credibility of the person or page recirculating the image became even more important.

 

Correction as Relational Authority

 

Discussion of verification practices highlights a second theoretical linkage: The social conditions of correction matter as much as message content. Meta-analytic work finds that corrections usually reduce misperceptions, but effects vary by source credibility, message design, and audience predispositions (Walter & Tukachinsky, 2020). Research on accuracy prompts suggests that shifting attention toward accuracy can reduce the sharing of misinformation (Pennycook et al., 2021). Corrective messages that matched the recipient’s primary language and invoked a locally legible authority, such as respected journalists, religious figures, elders, or trusted group administrators, were treated as less face-threatening and more actionable. This helps clarify why conflict avoidance, relationship maintenance, social identity, and local norms of authority are not peripheral to the findings. They are mechanisms through which misinformation and correction are socially evaluated. In identity-fragmented publics, accepting a correction can imply more than updating a factual belief; it can signal loyalty, disloyalty, suspicion, or trust. Corrections therefore worked best when they allowed users to withdraw support from a false claim without openly humiliating relatives, challenging respected figures, or appearing to side with an opposing group.

 

This aligns with evidence that conflict avoidance in interpersonal messaging environments suppresses people’s willingness to challenge misinformation, even when they doubt it (Chadwick et al., 2024). At the same time, the interview and survey evidence suggests a more nuanced pattern than simple suppression. In some cases, relationship maintenance reduced public endorsement, but redirected circulation into private and selective forwarding. That distinction is important in conflict-sensitive contexts because apparent declines in public sharing may mask continued interpersonal circulation. Correction was therefore relational as well as informational: It became credible when it came from someone socially authorized to speak, and it became shareable when it preserved the recipient’s identity and relationships.

 

Theoretical Contribution and Implications

 

The study also speaks to scholarship on African platform governance and linguistic inequality by showing how moderation gaps intensify the risks of multilingual drift. Research on connectivity initiatives and platform power in Africa highlights how corporate infrastructure and uneven access can reshape what the public sees and how they interpret it (Nothias, 2020). Ethiopia’s case adds that even when access expands, the safety and integrity of information are constrained by uneven investment in local languages and by the difficulty of moderating code-switched or low-resource language content. Recent work on Ethiopia argues that platform moderation can fail in part because of language blind spots and limited capacity to address harmful content across major local languages (Ayalew, 2025). However, the findings suggest that the problem is not only technical detection. It is also social fit: Institutional corrections often lose force when they arrive without the language, messenger, or relational cues that make them credible within a given public.

 

Overall, the study’s contribution is to show that misinformation transmission in multilingual publics cannot be fully explained by virality, platform design, or belief correction in isolation. In the Ethiopian case, multilingual transmission is organized through bridge actors, screenshot-based evidentiary forms, and socially embedded repertoires of trust that shape both amplification and correction. The broader theoretical contribution is to extend the two-step flow theory beyond a relay model of opinion leadership. In multilingual networked publics, opinion leadership operates through translation, platform brokerage, and relational authorization. Bridge actors are not simply intermediaries between media and audiences; they are interpreters who reshape what a claim means, who it appears to threaten, and which authorities can credibly correct it.

 

Concluding Remarks

 

This article shows that multilingual misinformation in Ethiopia does not merely spread across publics, but is actively remade through translation, code switching, and format conversion as content moves among Amharic, Afaan Oromo, Tigrinya, and Somali spheres. The broader theoretical implication is that multilingual misinformation should be understood as mediated reauthorization: Claims become powerful when bridge actors translate them into the language, identity frame, and authority structure of another public. The findings underline that cross-language diffusion is not platform-neutral. It is shaped by infrastructures and attention systems that privilege a small number of dominant gateways, with Telegram-style broadcast channels, repost cultures, and Facebook-page ecosystems enabling rapid relay and reframing at scale. Corrective messages were most persuasive when they matched the language of circulation, acknowledged locally salient threat perceptions, and came from actors with recognized moral authority in the relevant community. Platforms and regulators should treat linguistic capacity as a safety requirement rather than an optional localization feature, investing in moderation coverage for Ethiopian languages and code-switched content. More broadly, the findings suggest that multilingual low-resource contexts require not only more moderation but also different moderation strategies that are attentive to translation drift, screenshot-based circulation, and the uneven credibility of institutional messaging across segmented publics. Future studies should combine multilingual qualitative tracing with ethically grounded digital trace approaches such as data donation or screen-based logging.

 

 

References

 

Ababu, T. M., Woldeyohannis, M. M., & Getaneh, E. B. (2025). Bilingual hate speech detection on social media: Amharic and Afaan Oromo. Journal of Big Data, 12(1), Article 30. https://doi.org/10.1186/s40537-024-01044-y

 

Allen, J., Howland, B., Mobius, M., Rothschild, D., & Watts, D. J. (2020). Evaluating the fake news problem at the scale of the information ecosystem. Science Advances, 6(14), eaay3539. https://doi.org/10.1126/sciadv.aay3539

 

Ayalew, Y. E. (2025). Linguistic blind spots in platform content moderation in Ethiopia: The case of Facebook. Modern Languages Open, 2025(1), Article 11. https://doi.org/10.3828/mlo.v0i0.468

 

Ayalew, Y. E., & Ayalew, A. B. (2024). On selfies and hashtags: Disinformation during armed conflict in Ethiopia. In T. Roberts & G. H. Karekwaivanane (Eds.), Digital disinformation in Africa: Hashtag politics, power and propaganda (pp. 79–98). London, UK: Zed Books. https://doi.org/10.5040/9781350319240.0011

 

Behr, D. (2017). Assessing the use of back translation: The shortcomings of back translation as a quality testing method. International Journal of Social Research Methodology, 20(6), 573–584. https://doi.org/10.1080/13645579.2016.1252188

 

Byrne, D. (2022). A worked example of Braun and Clarke’s approach to reflexive thematic analysis. Quality & Quantity, 56(3), 1391–1412. https://doi.org/10.1007/s11135-021-01182-y

 

Chadwick, A., Vaccari, C., & Hall, N. A. (2024). What explains the spread of misinformation in online personal messaging networks? Exploring the role of conflict avoidance. Digital Journalism, 12(5), 574–593. https://doi.org/10.1080/21670811.2023.2206038

 

Chau, D., & Lee, C. (2021). “See you soon! ADD OIL AR!”: Code-switching for face-work in edu-social Facebook groups. Journal of Pragmatics, 184, 18–28. https://doi.org/10.1016/j.pragma.2021.07.019

 

Fiesler, C., & Proferes, N. (2018). “Participant” perceptions of Twitter research ethics. Social Media + Society, 4(1), 1–14. https://doi.org/10.1177/2056305118763366

 

Ford, E., Shepherd, S., Jones, K., & Hassan, L. (2021). Toward an ethical framework for the text mining of social media for health research: A systematic review. Frontiers in Digital Health, 2, Article 592237. https://doi.org/10.3389/fdgth.2020.592237

 

Golder, S., Ahmed, S., Norman, G., & Booth, A. (2017). Attitudes toward the ethics of research using social media: A systematic review. Journal of Medical Internet Research, 19(6), e195. https://doi.org/10.2196/jmir.7082

 

Grinberg, N., Joseph, K., Friedland, L., Swire-Thompson, B., & Lazer, D. (2019). Fake news on Twitter during the 2016 U.S. presidential election. Science, 363(6425), 374–378. https://doi.org/10.1126/science.aau2706

 

Haile, J. M. (2024). Social media for diffusion of conflict & violence in Ethiopia: Beyond gratifications. International Journal of Educational Development, 108, Article 103063. https://doi.org/10.1016/j.ijedudev.2024.103063

 

Herasimenka, A., Bright, J., Knuutila, A., & Howard, P. N. (2023). Misinformation and professional news on largely unmoderated platforms: The case of Telegram. Journal of Information Technology & Politics, 20(2), 198–212. https://doi.org/10.1080/19331681.2022.2076272

 

Inwood, O., & Zappavigna, M. (2024). The legitimation of screenshots as visual evidence in social media: YouTube videos spreading misinformation and disinformation. Visual Communication, 25(2), 446–472. https://doi.org/10.1177/14703572241255664

 

Katz, E., & Lazarsfeld, P. F. (1955). Personal influence: The part played by people in the flow of mass communications. New York, NY: Free Press.

 

Lazer, D. M. J., Baum, M. A., Benkler, Y., Berinsky, A. J., Greenhill, K. M., Menczer, F., Metzger, M. J., Nyhan, B., Pennycook, G., Rothschild, D., Schudson, M., Sloman, S. A., Sunstein, C. R., Thorson, E. A., Watts, D. J., & Zittrain, J. L. (2018). The science of fake news. Science, 359(6380), 1094–1096. https://doi.org/10.1126/science.aao2998

 

Mutz, D. C., & Young, L. (2011). Communication and public opinion: Plus ça change? Public Opinion Quarterly, 75(5), 1018–1044. https://doi.org/10.1093/poq/nfr052

 

Nieborg, D. B., & Poell, T. (2018). The platformization of cultural production: Theorizing the contingent cultural commodity. New Media & Society, 20(11), 4275–4292. https://doi.org/10.1177/1461444818769694

 

Nothias, T. (2020). Access granted: Facebook’s free basics in Africa. Media, Culture & Society, 42(3), 329–348. https://doi.org/10.1177/0163443719890530

 

Ohme, J., Araujo, T., Boeschoten, L., Freelon, D., Ram, N., Reeves, B. B., & Robinson, T. N. (2024). Digital trace data collection for social media effects research: APIs, data donation, and (screen) tracking. Communication Methods and Measures, 18(2), 124–141. https://doi.org/10.1080/19312458.2023.2181319

 

Orgeret, K. S., Mutsvairo, B., de Bruijn, M., Schroeder, D. T., Badji, S. D., & Moges, M. A. (2026). Hashtags, hatetags and social media campaigns in Ethiopia’s Tigray conflict. Information, Communication & Society, 29(3), 850–868. https://doi.org/10.1080/1369118X.2025.2533314

 

Pennycook, G., Epstein, Z., Mosleh, M., Arechar, A. A., Eckles, D., & Rand, D. G. (2021). Shifting attention to accuracy can reduce misinformation online. Nature, 592(7855), 590–595. https://doi.org/10.1038/s41586-021-03344-2

 

Poell, T., Nieborg, D., & Dijck, J. van (2019). Platformisation. Internet policy review: Journal on Internet Regulation, 8(4), 1–13. https://doi.org/10.14763/2019.4.1425

 

Pohjonen, M. (2022). An epistemic proxy war? Popular communication, epistemic contestations and violent conflict in Ethiopia. Popular Communication, 20(3), 236–252. https://doi.org/10.1080/15405702.2022.2074998

 

Quelle, D., Cheng, C. Y., Bovet, A., & Hale, S. A. (2025). Lost in translation: Using global fact-checks to measure multilingual misinformation prevalence, spread, and evolution. EPJ Data Science, 14(1), Article 22. https://doi.org/10.1140/epjds/s13688-025-00520-6

 

Roberts, J. K., Pavlakis, A. E., & Richards, M. P. (2021). It’s more complicated than it seems: Virtual qualitative research in the COVID-19 era. International Journal of Qualitative Methods, 20, 1–13. https://doi.org/10.1177/16094069211002959

 

Saunders, B., Sim, J., Kingstone, T., Baker, S., Waterfield, J., Bartlam, B., Burroughs, H., & Jinks, C. (2018). Saturation in qualitative research: Exploring its conceptualization and operationalization. Quality & Quantity, 52(4), 1893–1907. https://doi.org/10.1007/s11135-017-0574-8

 

Song, Y., Lu, Y., Tsang, S. J., Zhang, J., & Ku, K. Y. (2025). When corrections fail: Effects of misinformation targets, repeated exposure, and partisanship on misinformation beliefs. International Journal of Communication, 19, 392–416.

 

Swire-Thompson, B., Cook, J., Butler, L. H., Sanderson, J. A., Lewandowsky, S., & Ecker, U. K. (2021). Correction format has a limited role when debunking misinformation. Cognitive Research: Principles and Implications, 6(1), Article 83. https://doi.org/10.1186/s41235-021-00346-6

 

Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. G. Austin & S. Worchel (Eds.), The social psychology of intergroup relations (pp. 33–47). Monterey, CA: Brooks/Cole

 

Taylor, J., & Pagliari, C. (2018). Mining social media data: How are research sponsors and researchers addressing the ethical challenges? Research Ethics, 14(2), 1–39. https://doi.org/10.1177/1747016117738559

 

Toumaras, N. (2025a). Networked hyperlocal activists: Digital democracy and engagement in sub-Saharan Africa. Social Media + Society, 11(2). https://doi.org/10.1177/20563051251345945

 

Toumaras, N. (2025b). Refugee entrepreneurship as narrative practice: Identity, adaptation and embeddedness. International Journal of Entrepreneurial Behavior & Research, 1–20. https://doi.org/10.1108/IJEBR-04-2025-0453

 

Vosoughi, S., Roy, D., & Aral, S. (2018). The spread of true and false news online. Science, 359(6380), 1146–1151. https://doi.org/10.1126/science.aap9559

 

Walter, N., & Tukachinsky, R. (2020). A meta-analytic examination of the continued influence of misinformation in the face of correction: How powerful is it, why does it happen, and how to stop it? Communication Research, 47(2), 155–177. https://doi.org/10.1177/0093650219854600

 

Wang, D., & Yecies, B. (2024). Pandemic-incited intermediated communication | Intermediated communication via social media platforms during the COVID-19 pandemic—Introduction. International Journal of Communication, 18, 2828–2836.

 

Wilmot, C., Tveteraas, E., & Drew, A. (2021). Dueling information campaigns: The war over the narrative in Tigray. The Media Manipulation Casebook. https://casebook-static.pages.dev/case-studies/dueling-information-campaigns-war-over-narrative-tigray

 

Wilson, S. L., Lindberg, S., & Tronvoll, K. (2021). The best and worst of times: The paradox of social media and Ethiopian politics. First Monday, 26(10). https://doi.org/10.5210/fm.v26i10.10862

 

Workneh, T. W. (2019). Counter-terrorism in Ethiopia: Manufacturing insecurity, monopolizing speech. Internet Policy Review: Journal on Internet Regulation, 8(1), 1–22. https://doi.org/10.14763/2019.1.1394

 

Workneh, T. W. (2021). Social media, protest, & outrage communication in Ethiopia: Toward fractured publics or pluralistic polity? Information, Communication & Society, 24(3), 309–328. https://doi.org/10.1080/1369118X.2020.1811367

 

 

Copyright © 2026 (Ahmed Aynalem). Licensed under the Creative Commons Attribution Non-commercial No Derivatives (by-nc-nd). Available at https://ijoc.org.

https://doi.org/10.65476/y82ak581

 

 

 

[1] Conflict of interest: The author confirms there are no conflicts of interest pertaining to this work. Funding: The author did not receive financial support for the research, authorship, or publication of this work. Ethics approval: The study received ethics consultation from Jimma University, and all interview and survey participants provided informed consent.