Daily Archives: August 18, 2026

Authors: Paul ‘Seun Sosina, Odebiyi Olusola James

Abstract: Artificial intelligence is transforming national security governance by strengthening intelligence gathering, threat detection, border surveillance, cybersecurity, crime prevention and conflict early warning. However, its deployment across Africa raises concerns about privacy, algorithmic bias, accountability, misinformation, institutional readiness and technological dependence. This study examined artificial intelligence and national security governance in Africa through a comparative analysis of Nigeria, South Africa, Kenya, Rwanda and Egypt. Anchored in Technology Domestication Theory, the study adopted a qualitative multiple-case design. Data were obtained through documentary analysis of peer-reviewed studies, national policies, legislative frameworks, security strategies and African Union policy documents. The data were analysed using thematic synthesis and comparative policy analysis. Findings revealed that AI improves data processing, threat identification, surveillance capability and proactive security decision-making, although the scale of adoption varies across the five countries. Egypt and Rwanda demonstrate centralized policy arrangements, while Kenya and South Africa possess digital ecosystems but face implementation and regulatory challenges. Nigeria’s adoption is constrained by infrastructure deficiencies, institutional responsibilities and technical capacity. Shared challenges include inadequate oversight, poor data quality, skills shortages, algorithmic bias, privacy risks and dependence on foreign technologies. The study concludes that effective AI-supported security requires context-sensitive governance grounded in human rights, democratic accountability, strategic autonomy and human oversight. It recommends risk-based regulation, oversight, impact assessments, indigenous capacity development, stronger data protection and harmonized standards.

DOI: https://doi.org/10.5281/zenodo.21993615

Published by: vikaspatanker

Authors: Ritik Sharma, Balvir Singh Thakur

Abstract: General-purpose large language models (LLMs) are increasingly deployed as autonomous research agents: given a web-search tool and a task-specific instruction, they are expected to compile structured, factually complete documents from public sources with little human oversight. We report an empirical case study, spanning three successive system architectures of a deployed research-automation module, evaluating whether a GPT-5-class model can reproduce the completeness and accuracy of a curated reference document. We show that a completeness gap recurs at every layer of the pipeline, not just at the LLM layer intuition would blame first. At the LLM layer, a self-directed search-stopping criterion does not scale with true task size, producing near-complete coverage on a low-activity subject but only 9 of 16 known events (56%) on a high-activity subject under comparable search effort. On a large structured disclosure table, the model correctly reports the verified aggregate figure but declines to enumerate individual line items it cannot verify from search snippets, capturing 0% of line items in the case measured. Replacing the model's self-directed search with a deterministic, non-LLM pre-fetch of an authoritative source, combined with a mandatory reconciliation checklist, more than triples measured coverage (9 to 28 events) — but building that deterministic stage itself surfaced two further completeness bugs, both caught and fixed before production through direct re-verification against live source data: a naive entity-name query missing 4 of 5 known related records until query construction was corrected, and a silent 1,000-record pagination cap that would have discarded 69% of one subject's true history had it gone undetected. We then report the most severe instance of the same underlying pattern: in a substantially more deterministic third-generation architecture, purpose-built to eliminate LLM-driven table extraction entirely, a "most recently filed wins" version-selection heuristic silently selected an incomplete partial-amendment filing over a complete original filing for the same reporting period, capturing only 1 of 10 real holdings (3.8% of true portfolio value) — a bug caught only via comparison against a second, independent reference for a different subject, not by the team's own prior testing, and which in turn silently disabled a downstream data-quality safety check that depended on the same broken data. We report these as an empirical case study rather than a statistically powered benchmark, and specify the exact follow-up measurement needed to convert the paper's central open finding into a confirmed before/after result.

DOI: https://doi.org/10.5281/zenodo.21992676

Published by: vikaspatanker

Authors: Yathish Kumar S

Abstract: This paper proposes a theoretical synthesis between the predictive processing framework in computational neuroscience most rigorously formalized as Karl Friston's free energy principle and the literary phenomenon of narrative suspense. We argue that suspense is not merely a felt quality of certain texts but a measurable consequence of how narrative structures modulate prediction error across a reader's generative model of unfolding events. Drawing on empirical work in neuroaesthetics, psycholinguistics, and narratology, we develop a model in which classical plot devices foreshadowing, withheld information, dramatic irony, the ticking clock, and the twist function as deliberate engineering of precision-weighted prediction error. We review existing fMRI and behavioral evidence on narrative transportation and the Default Mode Network, and we present, as a Stage 1 Registered Report, a fully specified 2×2 within-subjects experimental design with power analysis, a pre-registered statistical analysis plan using linear mixed-effects modeling, and directional hypotheses derived independently from competing theoretical accounts. We close by addressing the principal objection to this research program that formalizing suspense in information-theoretic terms risks reducing literary value to a computable quantity and argue that this concern rests on a confusion between mechanism and meaning.

DOI: https://doi.org/10.5281/zenodo.21990599

Published by: vikaspatanker