Policy Significance Statement
Through a comparative analysis of discourse on the #NoMore movement, this study reveals significant disparities in focused topics and emotion across different geographic and linguistic lines. Local perspectives, particularly from Ethiopian users and content in the Amharic language, offer critical insights into local lived realities; however, they are often overshadowed by more amplified and dominant language discourse. These findings underscore the need for granular social media analysis that highlights underrepresented perspectives. Policymakers, mediators, and international actors must prioritize these diverse voices to develop strategies that reflect the complex realities on the ground, especially those from conflict-affected communities.
1. Introduction
In the current information age, the contestation of knowledge should extend to a wider scale, involving viewpoints from grassroots levels outside of the mainstream media. Such an approach would contribute to a more comprehensive understanding of diverse perspectives, enhance the accuracy of information by facilitating fact-checking and verification through users’ firsthand experiences, and also aid in reducing misinformation by offering alternative perspectives and counterarguments (Carilli, Reference Carilli2021). For instance, during the Syrian conflict, the public opinion among displaced Syrians in Lebanon suggests that the majority of Syrian refugees support the opposition, a large minority sympathizes with the government, and large numbers of Sunni Arabs also side with the government, challenging the narratives of the war from sectarians (Corstange, Reference Corstange2019). Similarly, Ezzina (Reference Ezzina2021) explored linguistic approaches in news discourse on the decade-long Palestine-Israel conflict. The analysis reveals ideological practices in prominent media news coverage, and the media constructs reality through opinions and propositions rather than facts.
Social media platforms have become popular arenas for discussions and information seeking, functioning as central hubs for knowledge exchange and sharing (Meilian et al., Reference Meilian, Thienmongkol and Nimnoi2022). Numerous studies have been conducted to understand users based on their posts on social media, a field often referred to as opinion mining (Dave et al., Reference Dave, Lawrence and Pennock2003). While social media users are not necessarily representative of the whole population of a country’s citizens, understanding the perspective and sentiments of those engaged participants is increasingly instrumental as it impacts the political landscape within and outside a country (Elefelious et al., Reference Elefelious, Getachew, Hailemariam and Moges2020). Examining their opinions through various lenses is vital to achieving a comprehensive understanding of social issues. This includes comparing and contrasting their viewpoints based on different background variables.
Some comparative studies have been done, but few focus on African issues. Several African nations encounter difficulties concerning political instability, violence, and ethnic tensions. International organizations and external countries actively engage in addressing these issues. Nevertheless, there is frequently a lack of awareness regarding the viewpoint of the local grassroots population on ongoing matters. This limited understanding manifests itself in various ways, including misrepresentation of community needs, amplification of dominant narratives, and the implementation of solutions that do not reflect the lived experiences and unique needs of these communities. Such outcomes not only reduce the effectiveness of interventions but also further marginalize already vulnerable populations. Understanding local people’s unique contexts and perspectives can pave the way for their voices to be represented in global discussions and shape decisions.
Some scholars have argued for the inclusion of African epistemologies when addressing African issues, rather than relying solely on frameworks and perspectives rooted in the “West,” typically referring to countries such as the US, Canada, Australia, New Zealand, and Europe (Kurth, Reference Kurth2003). For instance, Moyo and Mutsvairo (Reference Moyo, Mutsvairo and Mutsvairo2018) emphasize the need to rethink conceptual boundaries in Africa, urging for frameworks that reflect African realities and support continental integration. Pohjonen (Reference Pohjonen2022) in particular, studies the Ethiopian conflict that sparked the #NoMore movement and contends that the epistemic contestations within the hybrid media environment during this conflict are double-layered. They emphasize the importance of critically evaluating Western-origin theories and frameworks of knowledge, as these may not be universally applicable without contextual scrutiny. They seek to critically engage with the worldviews of those being studied alongside the researcher’s perspective, fostering our understanding of users from diverse perspectives without disregarding their information as irrelevant in epistemic decoloniality.
Although the concept of epistemic decolonization is complex and varies between disciplines (Long, Reference Long2023), the core idea challenges the colonial influence in knowledge production and academic research, highlighting the importance of inclusive perspectives and preserving diversity within the framework of universal knowledge (Knobloch, Reference Knobloch2020; Patin et al., Reference Patin, Sebastian, Yeon and Bertolini2020). It involves unlearning and dismantling unjust practices, assumptions, and institutions while simultaneously fostering positive actions and developing alternative ways of understanding (Mohamed et al., Reference Mohamed, Png and Isaac2020; Squire and Alozie, Reference Squire and Alozie2023). An approach to reshaping this dynamic is by amplifying the voices of marginalized communities on the global stage, developing theories rooted in their experiences, documenting their perspectives, and ensuring that their voices are heard (Duarte and Belarde-Lewis, Reference Duarte and Belarde-Lewis2015).
The #NoMore initiative began as a grassroots movement in November 2021. It was started and organized by the Horn of Africa Hub, a coalition of Ethiopian and Eritrean activists (cfeditoren, 2021). The primary goal of the movement is to raise awareness about the conflict in Ethiopia, oppose foreign interference, and demand an end to political and Western media misrepresentation. It stands against economic sanctions and highlights the distortion of local voices within international discourse (Staff, Reference Staff2021). The #NoMore campaign gained traction across various social media platforms using hashtags #NoMore and #SayNoMore, and #EthiopiaPrevails. Like the digital activism of the #OccupyWallStreet and #BlackLivesMatter movements (Freelon et al., Reference Freelon, McIlwain and Clark2016; Wang and Caskey, Reference Wang and Caskey2016), it utilizes social media platforms to disseminate information and rally global support. The impact of the movement has been particularly visible among the Ethiopian diaspora, where it sparked greater awareness and provided a platform for expressing local perspectives (cfeditoren, 2021). By challenging dominant media narratives and amplifying Ethiopian voices, the campaign has contributed meaningfully to international conversations around conflict representation and sovereignty.
This study investigates how opinions and sentiments are expressed within the #NoMore movement on Twitter. Given the limited research that centers local voices, it adopts a comparative analysis to understand how narratives are constructed, shaped, and communicated by different user groups, with particular attention to variation across geographic and linguistic contexts. Accordingly, the research is guided by the following questions:
RQ1: How do focused topics, expressed emotions, and the influence of opinion leaders differ between Twitter users located in Ethiopia and those in the US?
RQ2: How do focused topics, expressed emotions, and the influence of opinion leaders differ between tweets composed in Amharic and those written in English?
To address these questions, the study employs a comparative design along two key dimensions. First, it analyzes the opinions and emotional expressions of users in Ethiopia and the US. The US was selected due to its historical and ongoing involvement in the Horn of Africa (Abbink, Reference Abbink2021), as well as the influential role of the Ethiopian diaspora in shaping political discourse (Gagliardone and Pohjonen, Reference Gagliardone, Pohjonen and Mutsvairo2016). Second, the study explores the distinct perspectives of linguistic communities by examining tweets written in Amharic, one of Ethiopia’s official languages, and in English. By focusing on both language and location, the research adopts a context-sensitive lens to better understand locally grounded expressions and narratives. Theoretically, it draws on the concept of epistemic decolonization to examine the structural and discursive factors that contribute to divergences between local and global voices.
Our in-depth analysis of an Ethiopian political movement through tracking global and local narratives and identifying discrepancies in the movement’s core discussion topics and sentiments has the following contributions: It uncovers the differences in viewpoints between users from different regions and languages, offering an understanding of the viewpoints of local users on local conflicts that may have been overlooked in mainstream media. This enables local users’ voices to be heard in political decision-making, conflict resolution, and long-term peace-building efforts. Additionally, it documents and preserves views, knowledge, and strategies of local people, ensuring these are accessible for future studies and contributing to global knowledge production. It expands the decolonization of knowledge in communication concepts by leveraging new dataset analysis. Building upon existing frameworks and methodologies, our research seeks to incorporate fresh data to enhance the understanding of local people’s unique perspectives.
2. Literature review
2.1. #NoMore social movement
Ethiopia was ruled by the Ethiopian People’s Revolutionary Democratic Front (EPRDF), a coalition of regional parties dominated by the Tigray People’s Liberation Front (TPLF), which ruled the Tigray region and influenced Ethiopian politics from 1991 until 2018 (Getachew and Beshah, Reference Getachew and Beshah2019). On November 3, 2020, the Ethiopian federal government accused the TPLF of attacking an army base in Tigray, prompting the launch of a military offensive. The conflict continued until a ceasefire was signed on November 2, 2022. In November 2021, organizers from the Horn of Africa launched a social movement using hashtags #NoMore and #SayNoMore to oppose foreign interference, and #EthiopiaPrevails to highlight Ethiopia’s resilience and sovereignty (Korybko, Reference Korybko2022). The movement predominantly uses the #NoMore hashtag to spread its message, which centers on resisting international influence, disinformation campaigns, economic warfare, diplomatic propaganda, and military interventions. Additional hashtags such as #EthiopiaPrevails and #SayNoMore are used in parallel to reinforce and amplify this central theme. The hashtag #EthiopiaPrevails highlights Ethiopia’s resilience and determination to overcome challenges while asserting the country’s sovereignty and its right to make independent decisions (Borkena, Reference Borkena2021). The #NoMore movement leverages social media platforms. Facebook primarily serves for personalized interactions and news updates, whereas Twitter (now X) plays a pivotal role in political discourse in Ethiopia. Twitter has become the central platform for hashtag activism, as evidenced by previous local movements like #Oromo-Protests and #FreeZone9Bloggers (Gagliardone and Pohjonen, Reference Gagliardone, Pohjonen and Mutsvairo2016). The #NoMore initiative is a global social movement that addresses issues such as the misrepresentation of local voices, media distortion, political interference, and economic sanctions relevant to Africa.
2.2. Mining social media opinions on social movements
The influence of social media on political discussions has increased since its inception. Social media platforms have become essential tools for facilitating the exchange of information, influencing of opinions, and decision-making processes. With more than 368 million active users in 2022 (Statista, 2023), Twitter is one of the most influential social media networks in the world, particularly when it comes to political discourse. Numerous studies have analyzed political discourse by using data from Twitter. The #BlackLivesMatter, #MeToo, #BringBackOurGirls, and #IceBucketChallenge movements serve as prominent illustrations of how social media has evolved into a vital platform for political and social discourses (Schwörer and Fernández-García, Reference Schwörer and Fernández-García2021; Yu et al., Reference Yu, Mashhadi, Boy, Nielsen and Hong2022). Our study analyzes the #NoMore movement, which originated in Ethiopia and aims to surface the perspective of different user groups.
National boundaries and linguistic constraints do not appear to obstruct public online collective engagement on social media. Studies have compared the differences in the communication and engagement of users based on their locations (Alarcón-del-Amo et al., Reference Alarcón-del-Amo, Gómez-Borja and Lorenzo-Romero2015; Jackson and Wang, Reference Jackson and Wang2013). A comparison between US and South Korean social media users showed that ethical culture positively influences information-sharing behavior but negatively impacts concerns about information privacy (Chai, Reference Chai2020). Tsai and Clobert (Reference Tsai, Clobert, Cohen and Kitayama2019) demonstrated that culture shapes emotional life, and many assumptions derived from Western views of emotion may not fully apply to East Asian contexts. For example, Western cultures prioritize maximizing positive emotions and minimizing negative ones more than East Asian cultures do.
The rise of regional languages in digital platforms highlights the importance of linguistic diversity and inclusion in online conversations (Ndlangamandla, Reference Ndlangamandla2022; Gu et al., Reference Gu, Huang and Lin2024). Although English is the dominant language, other languages such as Spanish and Japanese are also widely used on Twitter, contributing significantly to the platform’s adoption (Hong et al., Reference Hong, Convertino and Chi2011). Language serves complex functions in shaping and constructing narratives on Twitter, emphasizing the advantages of tailoring research by language to gain deeper insights (Arazzi et al., Reference Arazzi, Nicolazzo, Nocera and Zippo2023). While language influences the popularity of content creators, it plays a less significant role in forming strong communities within online networks (Arazzi et al., Reference Arazzi, Nicolazzo, Nocera and Zippo2023).
In our study, we compare the voices from users in the US and local Ethiopian voices and those in English and Amharic languages by investigating variations in content, sentiments, and emotional responses.
2.3. Epistemic decolonization
Epistemic decolonization refers to the effort to challenge, dismantle, and reconfigure dominant Western knowledge systems that have historically marginalized non-Western and indigenous epistemologies (Mignolo, Reference Mignolo2009; Spivak, Reference Spivak2023). It promotes the recognition and inclusion of multiple ways of knowing, especially those rooted in local contexts, oral traditions, community-based knowledge, and indigenous languages (Green, Reference Green1999).
In the context of digital social media platforms, English discourse tends to dominate, both in terms of visibility and perceived legitimacy, often obscuring knowledge communicated in low-resource languages, such as Amharic in this study. This marginalization reflects a broader epistemic injustice, where what counts as credible, relevant, or “global” knowledge is shaped by Western norms and colonial legacies (Hookway, Reference Hookway2010). Furthermore, large amounts of data collected from social media platforms are often analyzed using Western-centric models, which may overlook alternative logics of meaning-making, emotional expression, and collective identity formation (Couldry, Reference Couldry2025).
By adopting the epistemic decolonization process as a guiding framework, this study aims to uncover and validate locally rooted expressions of opinion, emotion, and crisis response among Ethiopian Twitter users and Amharic language communicators. Specifically, the framework will be applied to examine whether disparities in power or dominance exist between local and global voices, and to determine whether perspectives shared by local users or expressed in Amharic offer distinct narratives.
3. Data collection
We gathered a dataset comprising 1,036,111 tweets by conducting searches using the hashtags #NoMore, #EthiopiaPrevails, and #SayNoMore. We collected the data using the Twitter (X) API for free, but shortly after completing our data collection, Twitter (X) announced the end of free API access. This data collection spanned from November 21, 2021, to August 21, 2022, encompassing the initial phase of the conflict in Ethiopia. These were among the most prevalent hashtags used in the #Nomore movement. We utilized the Tweepy package in Python to gather tweets (Wisdom and Gupta, Reference Wisdom and Gupta2016). We then removed tweets from bots (Yang et al., Reference Yang, Ferrara and Menczer2022) and mainstream media accounts, and retweets, to focus on individual users. We further processed the resulting dataset to remove punctuation, contractions, links (i.e., URLs), and stop words, and next lemmatized the data.
The dataset consisted of tweets written in multiple languages: 78% in English, 12% in Amharic (an official working language of Ethiopia), 7% in unspecified languages, and 3% in other languages. Among these, Amharic is the only Ethiopian language represented in the dataset and second only to English in frequency. Amharic language served as the only official language until 1991 and remains the most widely spoken language in terms of the total number of speakers in Ethiopia (Ethnologue, n.d.).
To facilitate comparative analysis, we used English-language tweets for the location-based analysis, comparing tweets from users located in Ethiopia and the US. In the language-based analysis section, we shifted focus to examine differences between tweets written in English and Amharic. Given the dominance of English in the dataset, we curated a subset to facilitate a more balanced comparison. The final sample consisted of 19,031 tweets, comprising 90% (17,130 tweets) in English and 10% (1901 tweets) in Amharic.
To enable direct comparison of discussion topics, emotional expressions, and influential users across the two languages, we translated the 1901 Amharic tweets into English using Azure Cognitive Translator. Human evaluators then assessed the translation quality, assigning a score of 1 for accurate translations, 0.5 for partially accurate, and 0 for inaccurate translations. The evaluation indicated an overall translation accuracy of 91%, demonstrating the tool’s effectiveness in translating Amharic tweets into English.
4. Methodology
4.1. Identification of user location
Following Lyu et al. (Reference Lyu, Fan, Xiong, Komisarchik and Luo2021), we used the location information in each user’s profile to identify their location. Geopy was used to convert location data into geocoded city, state, and country locations.
Noisy location information such as “My parent’s basement” was given a null value. Of the 351,882 tweets, 47.23% indicated African countries, primarily Ethiopia (87.58% = 146,213), as their location; 48.80% indicated countries in the Western world, primarily the US (73.27% = 125,838), Canada, Australia, New Zealand, and countries in Europe; and the other 3.75% indicated countries in Asia. It is plausible to assume that among these participants outside of Ethiopia are Ethiopian immigrants residing in different parts of the world interested in the situations unfolding in their home country. Furthermore, this may encompass citizens of other countries, activists, researchers, and organizations actively involved in addressing societal issues worldwide.
Our geolocation-based comparative analysis enables people in the US to access information from diverse perspectives rather than relying solely on mainstream media. Without this geolocation analysis, we would miss the differing perceptions of the movement, leading to a biased understanding of how people view issues in Ethiopia. This influence is reciprocal; African immigrants, shaped by their experiences in the host country, also have their voices heard by local people, thus impacting local decision-making and perceptions.
4.2. Topic modeling
Latent Dirichlet allocation (LDA) was used to extract topics and important words describing topics (Grootendorst, Reference Grootendorst2022). It used the sentence-transformers DistilBERT package with default parameters. LDA was implemented in two distinct phases. In the first phase, a location-based analysis was conducted using the complete dataset of 351,882 tweets containing location metadata. The analysis focused specifically on tweets originating from users in Ethiopia and the US. Together, the two countries accounted for over 77% of the location-tagged tweets, with 41.55% originating from users in Ethiopia and 35.76% from users located in the US. This subset yielded ten optimal discussion topics or clusters. In the second phase, a language-based analysis was performed on a sample of 19,031 tweets composed in both English and Amharic. From this dataset, five optimal topics were identified. Following topic modeling, each tweet in each cluster has tweets with associated probability scores. Two research assistants independently reviewed the top tweets with the highest probability scores in each group and extracted summaries and descriptive titles for each identified topic category.
4.3. Sentiment detection
To detect the emotions from tweets, we used a pre-trained deep learning model built on the DistilRoBERTa architecture that is designed to classify text into seven emotion categories. This distilled and efficient version of RoBERTa was pre-trained on a dataset with texts from Twitter and Reddit. The model effectively detects emotions in tweets (Hartmann, Reference Hartmann n.d.; Sanh et al., Reference Sanh, Debut, Chaumond and Wolf2019). The emotion category is based on Ekman’s definitions of the six basic emotions, including Joy, Sadness, Anger, Fear, Surprise, and Disgust (Ekman, Reference Ekman1992). The model first tokenizes the tweets, then runs predictions and calculates a score to categorize the text into specific emotion categories.
We manually assessed the accuracy of emotion detection by having two student assistants label emotions in a random sample of 500 tweets. The agreement rate of the labeling is 89%, and the Cohen’s kappa is 0.77, indicating a substantial agreement. Although the accuracy of emotion detection was not ideal, in the aggregated analysis, it may not heavily affect the comparative analysis, considering that the errors might be evenly distributed.
5. Results
5.1. Descriptive statistics
Between November 2021 and August 2022, the number of tweets per month fluctuated from 7411 to 466,068, presumably in reflection of the intensity of the conflict on the ground. The number of tweets was high in November and December 2021 as the TPLF advanced from the Tigray region toward the capital city, Addis Ababa, approximately 580 miles away. However, in late December 2021, the federal government’s forces ultimately pushed the TPLF back to the Tigray region, leading to a subsequent decline in online discourse (Figure 1).
Figure 1. Weekly posts related to the #NoMore movement collected from November 21, 2021, to August 21, 2022.
5.2. Location-based analysis
5.2.1. Topics by users in the US and Ethiopia
We analyze the differences in focused topics by users in Ethiopia and the US, who take 77.31% of all users in the dataset, where 41.55% is from Ethiopia and 35.76% is from the US. This significant proportion provides ample data to achieve our stated research objective and conduct a reliable and substantive comparative analysis between tweets from these two locations. Two research assistants reviewed the topics generated by LDA and selected the main themes. We ended up with ten main themes.
Topic 1 (The cause of the war: 38,557 tweets) restate and discuss how the war between TPLF and the federal government started on November 3, 2020. This seems to clarify who initiated the conflict amidst counterarguments.
Topic 2 (The implication of external influence: 62,544 tweets) encompasses the implications of external influence within Ethiopia and the continent of Africa. Tweets disapprove of the action of countries outside Africa on Ethiopia and criticizes foreign economic and political sanctions against Ethiopia, such as H.R.6600 (Sen. Menendez, Robert [D-NJ], 2022).
Topic 3 (Atrocities: 18,361 tweets) focuses on atrocities committed in Uganda and conflict zones in Ethiopia. Tweets also note that the media communicates that the war is in the Tigray region, whereas the conflict was in the bordering regions.
Topic 4 (Human rights abuse: 16,258 tweets) represents a discussion of human rights violations, killings, and crimes committed against humanity in various parts of Ethiopia due to the conflict. Tweets also pointed out fake news by individuals and mass media about human rights abuses and killings to influence the international community.
Topic 5 (The capital city and war crime: 6449 tweets) discusses the various challenges the country of Ethiopia encounters in socio-political environments, such as the diversity of the Ethiopian population. It rebuts the claim that the conflict happened because of the population’s diverse identity.
Topic 6 (Killings in Uganda: 9612 tweets) This topic discusses killings and abuse of innocent civilians in Uganda. Although the issue in Uganda has no link with the war in Ethiopia, users started to use #NoMore to reveal issues in their respective countries.
Topic 7 (Outsider intervention: 29,819 tweets) represents tweets focused on protesting countries outside the African continent interfering and meddling in Ethiopia’s internal affairs.
Topic 8 (Misinformation: 10,593 tweets) focuses on the disinformation propagated by mainstream media and some organizations about the conflict. The media was criticized discussing the conflict in a biased manner.
Topic 9 (Sovereignty: 47,996 tweets) focuses on defending and safeguarding the sovereignty of Ethiopia from foreign aggression and incursion as an independent ancient state.
Topic 10 (Unity of Ethiopia: 31,862 tweets) advocates for country unity since there are several ethnic-based tensions in Ethiopia. Tweets also appreciated users supporting #NoMore movement.
Tweets from Ethiopia and the US showed a comparable level of engagement in many topics, as shown in Figure 2. Reports by Borkena (Reference Borkena2021) have shown that global protests advocating the #Nomore movement were organized by Ethiopian and Eritrean communities or diasporas and delineated the active involvement of immigrants globally. The US users involved in the #Nomore movement were highly likely to be originally from Africa, particularly Ethiopia. Numerous studies highlight the active role of diaspora immigrants in their home countries’ politics. Lyons et al. (Reference Lyons, Ege, Aspen and Shiferaw2007) explored the influence of the Ethiopian diaspora in the US during the 2005 political crisis. Alebachew (Reference Alebachew2018) advocated for extending voting rights to Ethiopians abroad, citing their historical involvement, and political and economic reasons.
Figure 2. Volume of #NoMore posts by topic and location.
In Topic 1 (The cause of the war), Topic 6 (Killings in Uganda), Topic 7 (Outsider intervention), and Topic 9 (Sovereignty), tweets from Ethiopia, showed higher engagement than those from the US. Those topics focused on Ethiopian and combat-related issues, and Ethiopian users might have been more interested in discussing events they were experiencing and in proximity to (He et al., Reference He, Hong, Frias-Martinez and Torrens2015). By contrast, Topic 10 (Unity of Ethiopia) was more supported by users from the US (Figure 2). That difference shows the inclination and wishes of users in foreign countries to see a unified, peaceful Ethiopia probably because they reside in relatively peaceful countries and experience the benefits of doing so.
Furthermore, a significant difference in user engagement becomes apparent in Topic 2 (the implications of external influence) during specific weeks, as illustrated in Figure 3. Tweets from Ethiopia increased in engagement, specifically in Topic 2, starting in the first week of May 2022, when the U.S. Congress proposed the Ethiopia Peace and Stabilization Act of 2022 bill, S.3199, which would require imposing further sanction on Ethiopia on May 5, 2022 (Peace and of, 2022). Users showed increased engagement in requesting the bill’s cancelation, especially by using hashtags such as #CancelHR6600 and #CancelS3199.
Figure 3. Daily trajectories of #NoMore tweets proportions by topic, faceted by location.
5.2.2. Emotions by users in the US and Ethiopia
Across all topics and times, “anger” and “fear” consistently emerged as the dominant emotions. In contrast, “surprise” and “disgust” were the least frequently expressed (Figure 4). Users from Ethiopia consistently exhibited higher levels of “fear” than their counterparts from the US. Users from the US demonstrated more “sadness” compared to those from Ethiopia. This observed divergence in emotional expression may be attributed to the impact of cultural influences and societal expectations. Individuals may feel an elevated sense of fear, pride, anger, or indignation based on the narratives and interpretations circulated within their social circles. These emotional responses can shape their attitudes, behaviors, and willingness to engage in collective action (Fang et al., Reference Fang, Liu and Kawakami2023). A more comprehensive investigation is needed to analyze the socio-cultural factors influencing emotional expression within the #NoMore movement.
Figure 4. Weekly trajectories of the proportions #NoMore tweets by emotion (color) and location (line type).
Ethiopian users displayed a slightly elevated sense of “joy” following the week of the 16th, which coincided with the declaration of a humanitarian truce on March 24, 2022 (Figure 4). This truce reduced violence and crimes on the ground, consequently enhancing positive emotions among Ethiopian users.
There was a noticeable spike in “anger” from users in the US in the week of the 18th (Figure 4). This spike corresponded with a time when the U.S. Congress revisited the agenda of imposing further sanctions on May 5, 2022, through the Ethiopia Peace and Stabilization Act of 2022 (Peace and of, 2022).
5.2.3. Influence of users in the US and Ethiopia
As retweeting continues to play a crucial role in the distribution of information through social media, we employed an analysis of retweet counts as a key metric to gauge the extent of information dissemination and engagement (Firdaus et al., Reference Firdaus, Ding and Sadeghian2021). We focused on retweets as they are a more active form of engagement, indicating a user’s intent to amplify a message to their network or group. While “likes” indicate approval and comments provide qualitative feedback, “retweets” contribute directly to the spread and visibility of content across the platform. Focusing on individual users’ retweet counts allows us to measure the influence and reach of specific content and users directly.
We examined the volume of retweets received by specific tweets and the associated users responsible for originating those tweets. There were notable differences in information diffusion between Ethiopia and the US.
Tweets by users in Ethiopia received 624,245 retweets, while tweets by users in the US surpassed this with 1,280,697 retweets. Given the long-tail distribution, where few tweets receive many retweets, and most receive none or few, we used the Wilcoxon rank-sum test. Tweets from the US (n = 125,835) had significantly more retweets than those from Ethiopia (n = 146,213), with p = 0.00, showing that users from the US have more influence on the movement.
We generated the ECDF plot for retweet counts of fewer than 50 and those more than 50 (Figure 5). For retweet counts under 50, there’s little difference between tweets by Ethiopia and the US users. Yet, there are more originating from US users with counts surpassing 50, suggesting that influential voices are more from the US than from Ethiopia.
Figure 5. Empirical cumulative distribution of retweet counts for English tweets from Ethiopia and the US; the y-axis indicates the cumulative proportion of tweets at or below each value.
We further adopted the h-index, a metric that has been commonly used to measure the impact of academic scholars in their publications, to investigate influential users. A user has an h-index of n if he has at most n tweets that have been retweeted at least n times. Figure 6 shows the graph of the h-index of users from Ethiopia and the US. Both groups include influential users with an h-index higher than 100. However, users from the US group have a significantly higher h-index based on a Wilcoxon rank-sum test (p = 0.00). This further implies the dominant influence of voices from the US in the Ethiopian #NoMore movement. The higher h-index of influence in US users may reflect the systematic inequalities and biases of social media representations of US, possibly because US users have better economic resources and higher digital literacy, and collaborative networks and language advantages to spread their words, and algorithmic dominance on the platform (Hale, Reference Hale2014; Campos-Castillo and Laestadius, Reference Campos-Castillo and Laestadius2020).
Figure 6. The distribution of retweet h-index across the top 100 Twitter accounts ranked by retweet h-index from the US and Ethiopia.
When it comes to user influence, both Ethiopian and US groups have influential users, but there’s a distinction: Ethiopia’s top users include political figures like Prime Minister Abiy Ahmed Ali (@AbiyAhmedAli), Press Secretary Billene Seyoum (@BilleneSeyoum), and Ambassador Fitsum Arega (@fitsumaregaa). In contrast, the US group features journalists and activists like Hermela Aregawi (@HermelaTV), @mamamesay, @EdenCheckol, @wyattreed13, @NeaminZeleke, and @EADevCouncil, who are key figures in the #NoMore movement.
5.3. Language based analysis
5.3.1. Topics in English and Amharic
We analyzed tweets in English and Amharic to determine whether users express different opinions and emotions or have different opinion leaders in their native language compared to the dominant English language on social media platforms.
Tweets in Amharic primarily focused on issues within Ethiopia. Topic 1 called for an end to racism and extremism between different ethnic groups in the country, emphasizing unity and peace. Topic 2 highlighted concerns over killings, particularly the targeting of ethnic Amharas in Wollega and Benishangul, condemning violence and genocide. Topic 3 expressed frustration with government actions while discussing the impact of the third filling of the GERD (The Grand Ethiopian Renaissance Dam). Topic 4 focused on defending the current government’s power while condemning external influence and biased action against GERD progress. Topic 5 discussed various tensions and challenges in the country, urging more government action to end harmful actions and attitudes. These findings suggest that local-language tweets serve as a valuable source of information for understanding events at the national level. Conversely, the analysis of English-language tweets across five topics closely aligns with the 10 discussion themes in Section 5.2.1. It focused on casting a voice against external influence and misinformation about the ongoing conflict. Table 1 presents a summary of the five identified topics along with selected sample tweets in Amharic and their corresponding English translations.
Table 1. Amharic text discussion topics
5.3.2. Emotions in English and Amharic
The comparative analysis of emotions (Figure 7) revealed notable differences between Amharic and English tweets. Overall, Amharic tweets exhibited higher emotional intensity across all categories except for “joy.” Conversely, English tweets displayed a distinct spike in joy, largely driven by users defending the government, highlighting positive developments, and promoting the #UnityforEthiopia movement. However, this surge in joy was absent in Amharic tweets. Instead, a sharp increase in sadness was observed in Amharic tweets during a specific week, coinciding with discussions about the re-evaluation of H.R.6600 (Ethiopia Stabilization, Peace, and Democracy Act) by the U.S. Congress, amplified through the #CancelHR600.
Figure 7. Weekly trajectories of the proportions of #NoMore tweets by emotion (color) and language (line type).
5.3.3. Influence of users using English and Amharic
Similar to the location-based analysis, we examined whether influencer impact differs in English and Amharic tweets. The analysis revealed that Amharic tweets tend to have a lower retweet volume and a lower H-index compared to English tweets (Figure 8). Specifically, Amharic tweets received a total of 39,577 retweets, whereas English tweets significantly surpassed this with 226,909 retweets during the same period. The Wilcoxon rank-sum test indicated a significant difference in tweet diffusion by language, as measured by retweets. Tweets written in English (n = 17,130) had significantly more retweets than those written in Amharic (n = 1901), with p = 0.00. This result suggests that English-language tweets have a greater influence on the movement in terms of spreading through retweets. These findings could be attributed to several factors, including the usability of English content, differences in audience reach, and variations in user engagement across linguistic groups.
Figure 8. The distribution of the retweet h-index across the top 100 Twitter accounts ranked by retweet h-index posting in English and Amharic.
6. Discussion
We analyzed differences in online engagement between tweets from Ethiopia and the US and tweets written in English and the Amharic language in the #NoMore movement. This analysis revealed a spectrum of opinions shaped by users’ geographical locations and language, broadening the understanding of knowledge originating from Africa in this context.
A comparative analysis of tweets from Ethiopia and the US showed similar engagement levels. There is a potential that these users are either Ethiopians or Ethiopian immigrants in the US or have some form of connection to Ethiopia, leading to a comparable level of engagement and participation in each discussion topic. This indicates shared public opinions among Ethiopians globally. Studies on transnational activism have highlighted the growing use of social media by diasporic communities to maintain political ties and influence discourse in their countries of origin (Diminescu, Reference Diminescu2008; Bernal, Reference Bernal2014). Stakeholders, including embassies, should engage with the diaspora for deeper insights into Ethiopian conditions. The analysis also revealed that communities affected by conflict are more eager to discuss local issues than US users, highlighting the need for targeted strategies that address the specific needs of war-affected communities.
Tweets written in English and Amharic showed notable differences in the focus of topics. In Amharic tweets, users primarily discussed internal issues within the country, including regional conflicts, violence, and either praising or criticizing the government, which are largely directed to a local Ethiopian audience. In contrast, English tweets, similar to the topics identified in the location-based analysis, tend to focus on external influences, misinformation, and broader global concerns. Although there are some overlaps in topics between tweets written in Amharic and English, Amharic tweets raise additional concerns and perspectives specific to the country that may be overlooked in analyses dominated by English tweets. While this study did not include a formal civility assessment, Amharic-language conversations appeared to be more informal and exhibited a higher degree of incivility compared to their English tweets.These findings suggest that local governments and organizations can leverage language-based analysis to understand the community with inclusive perspectives and effectively address local concerns.
The emotional analysis revealed that “anger” and “fear” were the dominant emotions in tweets. Wollebæk et al. (Reference Wollebæk, Karlsen, Steen-Johnsen and Enjolras2019) shows that anger drives confirmation-seeking, while fear leads to exploring opposing views. Tweets by users in Ethiopia showed higher levels of “fear,” likely due to proximity to conflict, while tweets by US users, possibly from Ethiopian immigrants, expressed more “anger.” This contrast may reflect cultural differences, with Ethiopia’s collective culture fostering caution and the US’s individualistic culture encouraging freer expression.
From language perspective, Amharic tweets contained more emotional content than English tweets across most categories. One possible explanation could be that users articulate emotions more effectively in their native language. Another reason could be the differences in topics between the two languages, influencing the emotional variation. The prevalence of emotional expressions in Amharic texts may be attributed to several factors that warrant further investigation. One possible explanation is that the Amharic tweets originate from users within Ethiopia, whose first-hand experiences with ongoing events could contribute to heightened emotional intensity. Additionally, the linguistic characteristics of the Amharic language may facilitate more expressive or nuanced emotional articulation.
As retweeting stands as a pivotal avenue for disseminating information through social media, we found notable differences between users from Ethiopia and the US in information dissemination. Influential users from Ethiopia are mostly politicians, while those from the US are mainly journalists and activists.Differences in tweet diffusion and sharing paths also indicate variations in social circles between the two user groups. When comparing the influence of tweets in English and Amharic, we were unable to categorize opinion leaders’ profiles into specific categories, such as politicians or journalists, but we found that English tweets originated from accounts with a higher spreading capacity or influence compared to Amharic tweets. The dominance of English as the primary language on social media, along with the limited utilization of other languages, may have contributed to the difference in diffusion. Numerous studies highlight how English serves as a lingua franca on social media platforms, frequently marginalizing other languages in the process (Jackson, Reference Jackson2023; Lee, Reference Lee2020; Xue and Zuo, Reference Xue and Zuo2013).
By providing detailed insights into location- and language-based variations in discourse, this study offers a comprehensive understanding of public sentiment and social dynamics in the #NoMore movement. The significant differences in local and Western voices analyzed based on locations and Languages imply that the inclusion of voices from local communities is important. Voices from local users and language could be smaller in volume and drowned out by the dominant voices from US users or in English, but they provide a better understanding of local situations, reflecting what local people have experienced and their thoughts. In the #NoMore movement, the Ethiopian government may focus on Ethiopian voices to better address citizen concerns and local needs. Similarly, external countries and international organizations need to engage in fine-grained analysis to differentiate the needs of local people and the Western influence, using such insights to adjust their foreign policies and diplomatic relations. These analyses provide sources of information to validate on-the-ground realities, ensuring more informed decision-making. By incorporating granular public opinion analysis, governments and organizations can develop more inclusive policy frameworks, making their interventions and mediation strategies less biased and more representative of diverse perspectives. Furthermore, such analyses can support efforts to combat misinformation and disinformation about global issues, strengthening the integrity of international policies and responses.
7. Implication
From a theoretical perspective, this research highlights two primary implications. First, the core message of Ethiopia’s #NoMore movement, emerging in response to conflict, centers on national sovereignty and resistance to foreign intervention. This signifies a broader effort to counter external influence and misrepresentation while promoting endogenous knowledge production and self-determination, thereby directly aligning with the core tenets of epistemic decolonization. Second, geographically and linguistically based findings indicate that while users from Ethiopia and the US engaged with similar thematic areas, tweets originating in the US demonstrated significantly less emphasis on country-specific issues compared to those from Ethiopian users. This disparity suggests the potential marginalization or underrepresentation of local perspectives within international discourse. Furthermore, the analysis revealed linguistic marginalization: the predominance of English-language tweets contrasted sharply with Amharic-language content, which exhibited distinct thematic priorities, affective dimensions, and key influencers. These findings underscore the risk of epistemic exclusion inherent in analyses confined to dominant languages and highlight the critical necessity of integrating multilingual perspectives to foster inclusive and representative discourse, particularly during crises.
The results also underscore the urgent need for digital platforms to develop more robust tools for identifying, analyzing, and elevating content expressed in indigenous languages such as Amharic. Postcolonial digital studies argue that global platforms often marginalize non-Western epistemologies, reinforcing the need for inclusive, language-sensitive analysis in policymaking and international diplomacy(Couldry, Reference Couldry2025). Furthermore, this study points to the necessity of adopting decolonial digital methodologies that recognize and validate diverse epistemologies and ensure fair representation of marginalized linguistic and cultural groups in the development and training of artificial intelligence (AI) and large language models (LLMs). Without such efforts, valuable local knowledge risks being excluded from global digital discourse and automated systems that increasingly shape public understanding and decision-making.
Practically, the research highlights the need for global institutions, humanitarian organizations, and media platforms to adopt culturally informed, locally grounded decision-making processes and communication practices that move beyond universalized narratives rooted in dominant perspectives. The involvement of local actors and indigenous knowledge producers in shaping development policy and humanitarian response is essential to ensuring more inclusive, context-sensitive, and equitable global discourse. Furthermore, the findings provide valuable guidance for training future professionals in global communication, journalism, and conflict resolution, particularly in developing critical awareness of narrative bias and the importance of epistemic diversity.
8. Conclusion
Our study involved analyzing 1,036,111 tweets with the hashtags #NoMore, #EthiopiaPrevails, and #SayNoMore to answer the difference in online collective engagement between users from Ethiopia and the US, as well as between tweets in English and Amharic.
The results reveal that Ethiopian users on Twitter engaged more deeply with locally relevant topics, while U.S. users focused more on themes of peace and unity. Both groups frequently expressed anger and fear, though Ethiopian tweets showed higher levels of fear and occasional expressions of joy, while US tweets had notable spikes in anger during certain weeks. Additionally, their retweeting behaviors varied significantly. Correspondingly, tweets written in Amharic primarily centered on immediate domestic concerns and demonstrated stronger emotional intensity than those in English, which emphasized broader themes. Differences also emerged in how information was conveyed across the two language groups. Such comparative analysis based on user location and language is essential, as it highlights the importance of incorporating local voices that are often overshadowed by dominant mainstream narratives. This approach fosters the inclusion of diverse perspectives in global political discourse and helps document and preserve the views, knowledge, and strategies of local communities. Ensuring that these insights are accessible for future research contributes meaningfully to global knowledge production.
Our findings are limited by the small number of geo-tagged user data and the unverifiability of users’ location profiles. However, our analytical method is designed to minimize inaccuracies by including both location- and language-based analysis. Future research will compare social media opinions with corresponding news reports and interviews for more context, and explore information flow through network mapping to offer insights into group dynamics. Additionally, fine-tuning pre-trained machine learning models with African-specific datasets could improve their accuracy and relevance.
Data availability statement
The data supporting the findings of this study are openly available in the Ethiopian Social Movement #NoMore dataset (Zeleke et al., Reference Zeleke, Hong and Smith2025). For any additional inquiries regarding the dataset, please contact Meseret Zeleke at meseretzeleke@my.unt.edu and https://zenodo.org/records/15091380.
Author contribution
Conceptualization: M.Z., L.H.; Methodology: M.Z., L.H.; Data curation: M.Z.; Formal analysis: M.Z.; Data visualisation: M.Z.; Writing original draft: M.Z., L.H., D.S.; Writing review & editing: M.Z., L.H., D.S.; Funding acquisition: L.H.; Project administration: L.H., D.S.; All authors approved the final submitted draft.
Competing interests
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Ethical standard
The research meets all ethical guidelines, including adherence to the legal requirements of the study country.
Open access
Comments
No Comments have been published for this article.