Back to Stories

NOTABLE AI RESEARCH PAPERS - WEEKLY BRIEF #2026-8



February 23, 2026 - 3 min read

Financing the AI triad: compute, data and algorithms - A framework to build local ecosystems

Lucia Velasco, Sumaya Nur Adan, Muhammad Salar Khan, James Fox, Rodolfo Corona, Fola Adeleke, Jake Okechukwu Eoduh et al.

This paper proposes a staged financing framework for developing sovereign AI capabilities in low- and lower-middle-income countries, involving systematic investment across the 'AI triad' of data, computing power, and algorithms. Drawing on a comparative analysis of nine global funds in the areas of health, climate and infrastructure, the paper identifies five institutional design principles that distinguish effective financing mechanisms from failed initiatives.

Preserving historical truth: detecting historical revisionism in Large Language Models

Francesco Ortu, Joeun Yook, Punya Syon Pandey, Keenan Samway, Bernhard Schölkopf, Alberto Cazzaniga, Rada Mihalcea, Zhijing Jin

This paper introduces HistoricalMisinfo, a benchmark dataset comprising 500 contested historical events from 45 countries. Each event is paired with factual and documented revisionist narratives across 11 real-world prompt scenarios(questions, textbooks, social media posts, policy briefs). Using an LLM-as-a-judge evaluation protocol, the study reveals that, while models generally align with factual references under neutral prompts, they exhibit significantly higher revisionism scores when users explicitly request revisionist narratives revealing limited resistance to manipulative framing.

CreateAI: insights from an NSF Workshop on K-12 students, teachers, and families as designers of Artificial Intelligence and Machine Learning applications

Yasmin Kafai, José Ramón Lizárraga, R. Benjamin Shapiro

This NSF workshop report reimagines K-12 AI education, positioning students and teachers as creators and innovators of their own AI/ML applications, rather than merely competent users of AI systems. It addresses three critical questions: which tools, skills and knowledge empower students and teachers to build AI applications; how these approaches can be integrated into classrooms; and what new learning possibilities emerge from this creator-centred paradigm.

Human attribution of empathic behaviour to AI systems

Jonas Festor, Ivo Snels, Bennett Kleinberg

This research investigated whether people could distinguish empathetic qualities in relationship guidance produced by humans versus large language models through two preregistered studies involving 1,141 participants. Judgments were evaluated across quality and three empathy types: understanding others' perspectives, sharing feelings, and motivating action.

Agentic AI, medical morality, and the transformation of the patient-physician relationship

Robert Ranisch, Sabine Salloch

This paper examines how agentic AI systems may fundamentally alter the moral foundations of healthcare beyond familiar concerns of safety, accountability, and bias. Drawing on the framework of techno-moral change across three domains (decision, relation, and perception), the authors analyze how these increasingly independent AI agents might reconfigure the patient-physician relationship and transform core concepts of medical morality in ways that are not fully predictable.


Scan the QR code to view this story on your mobile device.


Agentic AIAI systemsCreateAI