Skip to content
Ndugelo Trevar PhophiAnthropology · Human Behaviour · AI

Portfolio

Case studies: what was asked, what was done, what changed.

A closer look at selected research and leadership work — the context behind each project, the approach taken and what it produced. For the full research programme, see Projects.

01 · Research

2026 — Honours dissertation, Nelson Mandela University

Algorithms of Trust

South African AI policy is written for a citizen who is assumed to understand, accept and benefit from automated systems. Public discussion online describes a different person entirely. The dissertation was formally presented to the Department of Sociology and Anthropology at Nelson Mandela University on 16 September 2026.

Approach

  • Assembled a corpus of public digital narratives about artificial intelligence in South Africa.
  • Applied Critical Discourse Analysis in Fairclough's tradition to the 2024 National AI Policy Framework and the withdrawn 2026 Draft AI Policy.
  • Compared the policy's imagined user against the self-descriptions of actual users.

Outcome

A working account of the imagined-user problem in South African AI governance, and a method other researchers can reuse to test whose experience a framework is actually designed around. The department responded positively to both the topic and the delivery, and faculty asked whether the line of research will continue into Master's study — a continuation now under consideration.

Presented to the Department of Sociology and Anthropology — 16 September 2026

Research themes

Sources & further reading

Collaboration

Reviewing the assumptions a policy or product makes about the people it is written for, and testing them against how those people actually describe the technology. How I work with organisations.

02 · Research

2025 — Undergraduate research strand, Nelson Mandela University

Digital trust in healthcare decision-making

Patients and clinicians rarely evaluate a digital health tool on accuracy alone. They evaluate whether someone will answer for the outcome. In a health system where access, language and continuity of care already vary widely, an accurate recommendation from an unaccountable source is not treated as reassurance — it is treated as one more thing to verify with a person.

Approach

  • Qualitative interviewing with students and health-service users about how they weigh advice from clinicians, family, search engines and AI-supported tools.
  • Thematic analysis of the accounts, coding for who is held responsible when a recommendation turns out to be wrong.
  • Mapped where trust attaches: to institutions and clinicians rather than to the system itself.
  • Tested the accounts against the consent and disclosure language used in digital health products.
  • Translated the findings into design implications for accountability, explanation and disclosure.

Outcome

A clearer articulation of trust as an institutional relationship rather than a property of a model: people extend trust to a system when a named person or institution remains answerable for its output. The strand now supplies the theoretical framework for the Honours work, and the interview and coding approach carried directly into Algorithms of Trust.

Research themes

Sources & further reading

Collaboration

Understanding why an accurate health or advisory tool is still not trusted, and where accountability and disclosure need to sit for people to act on it. How I work with organisations.

03 · Leadership

Nelson Mandela University student organisation

Ri Vha Venda — Acting President and Deputy Treasurer

A cultural student organisation needing steady governance, financial discipline and a programme that members would actually turn up to. Leadership had to hold together a membership that differs by language, home province and year of study, while remaining accountable to the university for every decision and every rand.

Approach

  • Led the executive as Acting President, chairing planning and representing the organisation to the university.
  • Managed budgeting and financial reporting as Deputy Treasurer, keeping records open to members.
  • Consulted members before committing the organisation, so decisions arrived with support rather than requiring defence afterwards.
  • Coordinated cultural and community events across a diverse student membership.
  • Handed over documented processes so the next executive did not have to rebuild them.

Outcome

Sustained membership engagement with transparent finances, and a governance record the next executive could take over intact — plus practical experience of the negotiation, consultation and accountability that the research writes about in theory. The same question runs through both: who is answerable, and do the people affected believe it.

Research themes

Sources & further reading

Collaboration

Governance, stakeholder consultation and decision-making in organisations that have to carry their members with them. How I work with organisations.

Back to the research programme