Charles Nimo, Tobi Olatunji, Abraham Toluwase Owodunni, Tassallah Abdullahi, Emmanuel Ayodele, Mardhiyah Sanni, Ezinwanne C. Aka, Folafunmi Omofoye, Foutse Yuehgoh, Timothy Faniran, Bonaventure F. P. Dossou, Moshood O. Yekini, Jonas Kemp, Katherine A Heller, Jude Chidubem Omeke, Chidi Asuzu Md, Naome A Etori, Aïmérou Ndiaye, Ifeoma Okoh, Evans Doe Ocansey, Wendy Kinara, Michael L. Best, Irfan Essa, Stephen Edward Moore, Chris Fourie, Mercy Nyamewaa Asiedu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Large language models (LLMs) are increasingly deployed in global healthcare, yet their outputs often reflect Western-centric training data and omit indigenous medical systems and region-specific treatments.This study investigates cultural bias in instruction-tuned medical LLMs using a curated dataset of African traditional herbal medicine.We evaluate model behavior across two complementary tasks, namely, multiple-choice questions and fill-in-the-blank completions, designed to capture both treatment preferences and responsiveness to cultural context.To quantify outcome preferences and prompt influences, we apply two complementary metrics: Cultural Bias Score (CBS) and Cultural Bias Attribution (CBA).Our results show that while prompt adaptation can reduce inherent bias and enhance cultural alignment, models vary in how responsive they are to contextual guidance.Persistent default to allopathic 1 (Western) treatments in zero-shot scenarios suggest that many biases remain embedded in model training.These findings underscore the need for culturally informed evaluation strategies to guide the development of AI systems that equitably serve diverse global health contexts.By releasing our dataset and providing a dual-metric evaluation approach, we offer practical tools for developing more culturally aware and clinically grounded AI systems for healthcare settings in the Global South.

While social computing technologies are increasingly being used to counter misinformation, more work is needed to understand how they can support the crucial work of community-based trusted messengers, especially in marginalized communities where distrust in health authorities is rooted in historical inequities. We describe an early exploration of these opportunities in our collaboration with Black and Latinx young adult "Peer Champions" addressing COVID-19 vaccine hesitancy the U.S. state of Georgia. We conducted interviews engaging them with a social media monitoring and outreach dashboard we designed, to probe their understanding of their roles and current and potential use of digital platforms. With the concept of cultural code-switching as a framing, we found that the Peer Champions leveraged their particular combination of cultural, health, and digital literacy skills to understand their communities' concerns surrounding misinformation and to communicate health information in a culturally appropriate manner. While being positioned between their communities and public health research and practice motivated and enabled their work, it also introduced challenges in finding (mis)information online and navigating tensions around authenticity and respect when engaging those close to them. Our research contributes towards characterizing the valuable and difficult work trusted messengers do, and (re)imagining collaboratively designed interpretive digital tools to support them.

Given Myanmar’s historical and socio-political context, hate speech spread on social media have escalated into offline unrest and violence. This paper presents findings from our remote study on the automatic detection of hate speech online in Myanmar. We argue that effectively addressing this problem will require community-based approaches that combine the knowledge of context experts with machine learning tools that can analyze the vast amount of data produced. To this end, we develop a systematic process to facilitate this collaboration covering key aspects of data collection, annotation, and model validation strategies. We highlight challenges in this area stemming from small and imbalanced datasets, the need to balance non-glamorous data work and stakeholder priorities, and closed data-sharing practices. Stemming from these findings, we discuss avenues for further work in developing and deploying hate speech detection systems for low-resource languages.

This paper describes co-located collaborative information seeking in the context of the Social Media Tracking Centre (SMTC) in Ghana. The SMTC was operational for three days during the Presidential elections in Dec 2016. The SMTC»s role was to collaboratively find and verify novel, actionable, and relevant information on social media and escalate it to the authorities to use that information to ensure a transparent and peaceful election process. We performed a qualitative analysis of semi-structured interviews with the volunteers at the SMTC as well as its managing team. We present, in this paper, the importance of volunteer motivation and co-location in the success of the SMTC, as well as the users» feedback on the collaborative tool, informing future design, derived from our analysis.

Since 2011, our research group, along with numerous local partners, has been building a platform and methodology for monitoring elections using social media. Historically, election monitoring has traditionally been the domain of trained monitors provided by international monitoring groups. But monitoring by domestic groups with fewer resources has been a growing phenomenon, supported in part by the availability of inexpensive digital technologies such as SMS. Social media represents a further, exciting step in this trend. We describe our five years of experience in this endeavor and report a series of key lessons learned. These lessons touch on issues such as source types and curation, collaboration with other election-related groups, human vs. automated analysis, varying stakeholder needs, and the value of falsification. We also share our vision for the next five years of this research.

Traditional election monitoring involves trained observers recruited and deployed by neutral organizations to complete structured reports based on their first-hand observations. In contrast, social media-based election monitoring takes as input the high volume, noisy data produced by individuals posting on social media outlets, with limited constraints and structure. Trained volunteers process this data in an effort to extract meaningful information. Both methods have been used successfully to support free and fair elections. In this work we begin to explore a middle ground, namely reports by untrained individuals, but mediated through a mobile phone application that provides structure for the responses. In collaboration with Pol-IT, a Buenos Aires based organization focused on politics and ICTs, we report on a first study of users of the Ojo con el Voto (Eyes on the Vote) application deployed during the Argentina presidential run-off election in November 2015. We expect that citizen apps for election monitoring will become increasingly popular, hence this early look at their use offers an opportunity to establish an initial baseline and to potentially influence subsequent development.

Social media offers new ways for citizens to discuss and debate politics and engage in the democratic process. These online systems could be places for rich policy relevant debate, which is favored by scholars of deliberative democracy. Alternatively, social media might be a platform for an identity driven form of political discourse that is routinely scorned by scholars of democracy. To examine these two possibilities, we analyzed tweets sent during three national elections, the defining participatory process of democracy. Our dataset includes over 760,000 tweets gathered during national elections in Nigeria, Ghana and Kenya from 2011 to 2013. In order to analyze the degree to which Twitter was being used for policy relevant discussion we developed policy term sets through a text analysis of the major political party platforms. To examine the amount of discourse focused on identity issues we created identity term sets based upon national religious, tribal, and regional differences. In Nigeria, where divisive identity politics feed violence and electoral misconduct, discussion of tribe, region, and religion dominate mentions of platform policies. In contrast Ghanaians, who enjoy the most robust democracy of the three countries, were seven times more likely to discuss policy issues rather than identity. Kenyan democracy is still undergoing consolidation, and tweets again reflect this, with almost as many tweets devoted to tribal identity as campaign policy. These findings suggest that social media discussions may echo the state of democratic deepening found in a country during its national elections.

Today is an exciting time to be a political activist in sub-Saharan Africa, particularly for the technically inclined. Online social media and other digital technologies are increasingly being used for political purposes. But this phenomenon raises the question of how, if at all, these new media actually perturb the political landscape. These questions have been well-studied in Western contexts, but remain relatively underexplored in developing regions where traditional media are often scarcer, democracies are younger, and the effect of social media on politics has the potential to be quite distinct. This paper explores these questions through a qualitative dual case study of social media use during general elections in Nigeria and Liberia in 2011. Participants suggested that social media helps to overcome previous scarcity of information during the electoral process, leading to increased transparency and reduced tension. Furthermore, social media-based monitoring shows encouraging signs of robustness concerning information quality and mobility. Together, these findings suggest that given sufficient civil-society coordination, social media can be an effective tool for electoral scrutiny and can help build public trust in the electoral process.

In this paper, we explore the different ways in which people collaborate and share knowledge in public Internet venues, or cybercafés, in Ghana, West Africa. Based on 150 survey interviews conducted in two different cybercafés, one urban and business-oriented and the other peri-urban and family-oriented, we find that most cybercafé customers, largely regardless of their demographic, would like to engage in collaborative work in public Internet venues and that a large percentage already are.

While the history of traditional media in post-conflict peace building efforts is rich and well studied, the potential for interactive new media technologies in this area has gone unexplored. In cooperation with the Truth and Reconciliation Commission of Liberia, we have constructed a novel interactive kiosk system, called MOSES, for use in that country's post-conflict reconciliation effort. The system allows the sharing of video messages between Liberians throughout the country, despite the presence of little or no communications infrastructure. In this paper, we describe the MOSES system, including several innovative design elements. We also present a novel design methodology we employed to manage the various distances between our design team and the intended user group in Liberia. Finally, we report on a qualitative study of the system with 27 participants from throughout Liberia. The study found that participants saw MOSES as giving them a voice and connecting them to other Liberians throughout the country; that the system was broadly usable by low-literate, novice users without human assistance; that the embodied conversational agent used in our design shows considerable promise; that users generally ascribed foreign involvement to the system; and that the system encouraged heavily group-oriented usage.

We examine the extent to which user-based subsidies can promote the sustainability and development impact of telecenters, where sustainability is defined in financial and social terms. We do this by looking at a coupon scheme used by the USAID funded eCenter network in Kyrgyzstan. The network consisted of partnerships with existing commercial computer centers which provided fee-based ICT services to their communities. The eCenter program temporarily provided subsidized coupons for Internet access and computer training to users of these centers.

Liberia is a country emerging from years of protracted and devastating civil conflict. Left without any fixed line telephone infrastructure, it relies solely on the mobile phone for telephony. This study investigates the usage of mobile phones in this immediate post-conflict setting. In particular, we adopt the uses and gratifications approach to media research, giving focus to both instrumental and intrinsic motivations for use. Mobile phone users in both the capital city of Monrovia and in various rural areas were surveyed using the Q methodology, which identified distinct perspectives within these urban and rural groups. Participants were then sorted into groups where each group contained users with similar perspectives on their mobile phones. These identified groups included sets of users who saw their phones as productivity enhancers, means of connectivity to family and friends, essential business tools, technological curiosities, and sources of personal security. The idea of a phone as a stylish object was markedly rejected, especially in rural areas. We contrast these Q-sort results from Liberia with previous work from Kigali, Rwanda, finding differences especially as related to security.

This paper presents Community Knowledge Sharing (CKS), an asynchronous discussion system designed for use in the developing world. CKS supports community development by allowing communities to interact, share and learn from each other. Recognizing that large segments of the population in rural communities have low levels of literacy, CKS implements a multi-literate design in which the system can be customized based on the abilities and preferences of the user. An evaluation of CKS, conducted in a rural agricultural community in the Dominican Republic, found that low literate users prefer fully iconic interfaces. Future systems designed for use by people with a range of literacy skills must ensure that iconic interfaces provide users with the audio and visual cues they need to work effectively.

Phenomenal improvements in the computational performance of multiprocessors have not been matched by comparable gains in I/O system performance. This imbalance has resulted in I/O becoming a significant bottleneck for many scientific applications. One key to overcoming this bottleneck is improving the performance of multiprocessor file systems. The design of a high-performance multiprocessor file system requires a comprehensive understanding of the expected workload. Unfortunately, until recently, no general workload studies of multiprocessor file systems have been conducted. The goal of the CHARISMA project was to remedy this problem by characterizing the behavior of several production workloads, on different machines, at the level of individual reads and writes. The first set of results from the CHARISMA project describe the workloads observed on an Intel iPSC/860 and a Thinking Machines CM-5. This paper is intended to compare and contrast these two workloads for an understanding of their essential similarities and differences, isolating common trends and platform-dependent variances. Using this comparison, we are able to gain more insight into the general principles that should guide multiprocessor file-system design.