Kabarak University

AfriVoices-KE is a large-scale multilingual speech dataset comprising approximately 3,000 hours of audio across five Kenyan languages: Dholuo, Kikuyu, Kalenjin, Maasai, and Somali. The dataset includes 750 hours of scripted speech and 2,250 hours of spontaneous speech, collected from 4,777 native speakers across diverse regions and demographics. This work addresses the critical underrepresentation of African languages in speech technology by providing a high-quality, linguistically diverse resource. Data collection followed a dual methodology: scripted recordings drew from compiled text corpora, translations, and domain-specific generated sentences spanning eleven domains relevant to the Kenyan context, while unscripted speech was elicited through textual and image prompts to capture natural linguistic variation and dialectal nuances. A customized mobile application enabled contributors to record using smartphones. Quality assurance operated at multiple layers, encompassing automated signal-to-noise ratio validation prior to recording and human review for content accuracy. Though the project encountered challenges common to low-resource settings, including unreliable infrastructure, device compatibility issues, and community trust barriers, these were mitigated through local mobilizers, stakeholder partnerships, and adaptive training protocols. AfriVoices-KE provides a foundational resource for developing inclusive automatic speech recognition and text-to-speech systems, while advancing the digital preservation of Kenya's linguistic heritage.

Music streaming platforms and short video services have become dominant intermediaries in the global circulation of African music, yet mounting evidence suggests they simultaneously reproduce structural disadvantages for African artists, genres, and audiences. This scoping review synthesises the interdisciplinary literature on how algorithmic systems, revenue models, and data governance regimes produce what we conceptualise as platformed invisibility: the systematic, opaque, and technologically mediated marginalisation of non-Western cultural expression on ostensibly universal digital infrastructures. Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) and Joanna Briggs Institute methodology, we searched seven bibliographic databases, targeted grey literature, and hand searched six leading journals for records published between January 2015 and June 2026. A total of 1,033 records were screened; 41 met inclusion criteria and were charted using an a priori extraction form. Narrative synthesis identifies four interlocking mechanisms, namely construct, method and item bias, algorithmic opacity, popularity feedback loops, and demographic and geographic skew, that operate within a political economy of data extractivism and infrastructural dependency. Countervailing interventions cluster around trustworthy artificial intelligence, data sovereignty, intellectual property reform, culturally grounded datasets, decolonial pedagogy, and Afrofuturist design imaginaries. The review advances three contributions: a conceptual definition of platformed invisibility that distinguishes it from adjacent constructs; a mapping of the mechanisms to interventions terrain; and a research agenda for empirical audits of African music on major streaming platforms. Findings underscore that algorithmic fairness for African music cannot be achieved through platform tweaks alone but requires structural decolonisation of technology, economy, and epistemology.

Generative artificial intelligence (AI) is reshaping strategic communication across Africa, yet the tools are trained overwhelmingly on Global-North data and carry its representational defaults. This paper asks whose voice such systems amplify, and whose they silence, when African organisations use them to speak to their own publics. Using qualitative documentary analysis of publicly reported cases in which African organisations used generative AI in outward-facing communication and drew criticism on representational or authenticity grounds, and applying reflexive thematic analysis through a decolonial, algorithmic-justice lens, the study constructs three themes. First, a double bind of representation: uncritical AI fails along two opposite axes, erasing the local subject by rendering it as Global-North, or essentialising it by reducing it to stereotyped African signifiers, both rooted in training-data geography. Second, a political economy of authenticity, in which the turn to AI displaces local creative labour and audiences read the resulting communication as inauthentic partly because they infer that displacement. Third, the public as algorithmic watchdog, whereby affected publics, not regulators, detect and contest AI use, performing a grassroots decolonial critique. A contrast case of a brand publicly refusing AI underscores that authenticity has become a contested market value. The paper offers a conceptual model of algorithmic representational justice and argues that just AI use in African strategic communication requires re-centring the displaced local voice, not merely improving model outputs. Cases cluster in Kenya and South Africa, taken here as an information-rich entry point into a wider Global-South dynamic.

This study develops and examines a sociotechnical framework for digital government implementation, focusing on how technical, organisational, and institutional conditions shape implementation outcomes in the public sector. Despite global investment in digital transformation estimated at over US $600 billion in 2024, failure rates for public sector information systems remain high, particularly in developing contexts. Existing research often treats technical, managerial, and behavioural factors separately. This study adopts an integrated perspective by conceptualising implementation as the interaction of technical factors such as system quality and interoperability, organisational factors including leadership, training, and change management, and institutional factors such as policy frameworks and governance structures. Within this framework, organisational resistance is conceptualised as an important explanatory process, while institutional capacity is treated as a relevant contextual condition that may shape implementation outcomes. The study draws on a convergent mixed-methods, multi-case design based on data from 153 survey respondents and 25 key informants across four Kenyan government ministries implementing electronic Government Procurement (e-GP) and Integrated Financial Management Information Systems (IFMIS). The analysis combines descriptive statistics, correlation analysis and qualitative thematic insights. The findings indicate that interoperability constraints (M = 3.2), uneven training (M = 3.3), and moderate institutional capacity (M = 3.5) are key implementation challenges.Correlation analysis shows that sociotechnical factors are positively associated with system performance (r = 0.56, p < 0.001) and negatively associated with organisational resistance (r = -0.41, p < 0.001). Organisational resistance is negatively associated with system performance (r = -0.48, p < 0.001), while system performance is positively associated with public value outcomes (r = 0.61, p < 0.001). These findings provide preliminary support for the proposed framework and highlight the importance of alignment across technical, organisational, and institutional domains. The study contributes to digital government research by offering a structured explanation of how implementation conditions are associated with performance and public value outcomes. It also provides practical guidance for policymakers, emphasising the need to strengthen interoperability, invest in user capacity, and treat resistance as a signal of underlying implementation misalignment.

Abstract Context Survival of low-birth-weight (LBW) neonates depends heavily on modifiable nursing care processes at the bedside, making the newborn unit a decisive site for improvement. Evidence linking measurable care-process conditions to outcomes, and the provider experience that explains them, remains limited in Kenyan county referral settings. Aim To examine the care-related determinants of severe adverse outcomes among LBW neonates and the provider perspectives that explain them, framed for newborn-unit nursing practice and quality improvement. Methods A convergent mixed-methods design was applied at Kericho County Referral Hospital. Quantitatively, 169 LBW neonate-mother pairs were analysed; care-process indicators (skilled personnel at admission, warm-chain maintenance, shortage of essential drugs/feeds, and referral/outborn status) were related to a composite severe adverse outcome using Pearson chi-square tests and crude odds ratios (OR) with 95% confidence intervals (CI). Qualitatively, key-informant interviews with newborn-unit providers were analysed thematically and coded to care-process themes; strands were integrated for practice. Findings A severe adverse outcome occurred in 136/169 neonates (80.5%). Skilled personnel (94.7%) and warm-chain practices (92.9%) were near-universal, whereas 58.6% of neonates faced shortages of essential drugs/feeds and 49.1% were referred (outborn). Shortage of essential drugs/feeds (OR = 2.26, 95% CI [1.04, 4.90], p = .036) and referral/outborn status (OR = 2.25, 95% CI [1.01, 5.00], p = .043) were significantly associated with higher odds of a severe outcome. Warm-chain care was statistically associated but in a counterintuitive direction (OR = 4.81, p = .006), consistent with confounding by indication, while skilled-personnel availability was not associated (p = .834). Provider narratives converged on seven care-process themes: staffing and workload, warm-chain maintenance, infection prevention, drug and equipment availability, referral coordination, monitoring and documentation, and caregiver/transport barriers around the first hour of care. Conclusion Care-related conditions—commodity supply, referral readiness, thermal care, infection prevention, and monitoring—are clinically modifiable levers that shape whether vulnerable LBW neonates stabilize or deteriorate, even where they do not all retain independent statistical significance after adjustment. Recommendations Newborn units should protect nurse staffing norms, secure consistent supply of essential neonatal commodities, standardize pre-referral stabilization and thermal-care protocols, and strengthen structured monitoring and documentation, supported by competency-based training and county-level policy.