Against Softmaxing Culture: Understanding Relational Practices in Expert and Ordinary Forms of Work @MicrosoftResearch
Against Softmaxing Culture: Understanding Relational Practices in Expert and Ordinary Forms of Work  @MicrosoftResearch
Uploaded July 2025 | Updated November 2026, 3 hours ago
What can human-AI interaction research learn from the practices of fundi wa simu (phone repair technicians)? In this talk, Daniel will show how everyday practices of collaboration and exchange are structured by multiple and overlapping norms, rules, values and expectations. He argues that effectively designing AI for such dynamic environments requires sustained attention to the complexities of culture and more grounded evaluation regimes.

Learn more about Microsoft Research Lab – Africa, Nairobi: microsoft.com/en-us/research/lab/microsoft-research-lab-africa-nairobi/seminars
Against Softmaxing Culture: Understanding Relational Practices in Expert and Ordinary Forms of WorkNeurIPS 2024: AI for Science with Chris BishopThe Illusion of Inclusion How LLMs Misrepresent African Languages and Cultural ContextsHow I became a StoryTeller (and how YOU can too)Learn the rules: The hacker mindsetARO: A new lens on matrix optimization for LLMsLanguage & Voice AI for Africa: From Data to Deployment and ImpactPromptionsMagenticLite: A full-stack agentic experience powered by Small ModelsWhat is Density Functional Theory (DFT)Abstracts: Zero-shot models in single-cell biology with Alex LuDeep Reinforcement Learning in Supply Chain Optimizations
Microsoft Research |

Against Softmaxing Culture: Understanding Relational Practices in Expert and Ordinary Forms of Work

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER