AI Ethics Consultant & Technology Specialist · Technical Writer · Journalist
Thirty years of building with technology. A career spent making it understandable. Now asking the questions that get skipped when everyone's moving fast.
I'm Mandy Hathaway, an AI ethics consultant and technical writer. Organizations hire me when they need someone who understands how these systems actually work and can communicate that clearly: to leadership, to regulators, to the public, or to their own teams. I hold an MA in Ethical Technology & Artificial Intelligence from Metropolitan State University and 30 years of hands-on technology experience.
My work sits at the edge where machine learning meets communication and accountability. I advise on responsible AI practices, write documentation that reflects how these systems actually work, and identify the ethical questions that get skipped when moving fast.
Past clients and employers include Coding With Kids, where I led a teaching team of up to 35 instructors, and The Metropolitan, where I served as Technology Editor for three years. Earlier in my career, I served at Pillsbury United Communities through AmeriCorps CTEP (the Community Technology Empowerment Project), developing curricula and technical documentation for their technology literacy programming. I've published on algorithmic bias, responsible AI development, and the gap between how AI systems are marketed and what they can actually do.
I consult and write remotely: AI ethics advisory, AI product consulting, technical writing for AI/ML products, prompt engineering, conversational design, and responsible AI documentation.
Core Expertise
Languages & Tools
Certifications & Coursework
Memberships & Affiliations
I build and evaluate the pieces that make AI systems trustworthy. RLHF and preference dataset design, annotation methodology, conversational agent design, model evaluation, and responsible AI policy review. My published work includes a human annotated AI ethics preference dataset and a DPO fine tuning notebook for ethics alignment. I bring the same rigor to client systems: bias and fairness audits, governance frameworks, and ethics training for non technical teams.
I build the infrastructure that lets small teams and solo operators put AI to real work: workflow automation, custom Claude plugins, MCP servers, scheduled task systems, and memory and skill architectures. I run my own consulting business on a system I built this way, so every piece of it is battle tested on a real workload before I hand it to a client. The same approach works for nonprofit staff drowning in repetitive intake, solo consultants who need a working second brain, and product teams building internal AI tooling.
I write documentation for AI and ML products that reflects how the systems actually work: API references, user guides, developer tutorials, prompt engineering libraries, and internal knowledge bases. I know how to read a codebase, talk to engineers, and produce documentation that people actually use. No jargon for its own sake. No oversimplification that misleads.
I design and deliver AI education for audiences who need the real picture: what these systems can and can't do, where they fail, and why the failure modes matter. My courses have reached students from elementary school through professional development contexts. I can also write plain language explainers and briefing documents for leadership and board level audiences.
Nonprofits face AI decisions without the legal, technical, and governance staff that enterprises take for granted. I work with mission driven organizations to vet tools before adoption, write policies the board can read and the staff can follow, and flag where AI helps the mission and where it risks the people the mission exists to serve. I hold the AI for Nonprofits Professional Certificate through Microsoft Elevate, LinkedIn, and NetHope.
Read the practical guide / One page summary
Download PDFs: Full guide (PDF) / One pager (PDF)
Some organizations need ongoing AI ethics oversight but don't have the volume to justify a full time hire. I work as a monthly retainer advisor for product teams, nonprofits, and editorial operations that want a thinking partner embedded in their decisions: reviewing launches before they ship, reading contracts with AI clauses, flagging risk early, sitting in on roadmap reviews, and keeping governance current as the laws keep shifting. A typical engagement runs four to eight hours a month with scoped access to your tools and team.
I write and think critically about the human dimensions of artificial intelligence, from bias and accountability to how we design systems that are genuinely equitable. Below is a selection of essays, research, and public writing on these questions.
AI systems don't generate bias from nowhere. They learn it from us. This essay traces how racial and gender inequity becomes encoded in machine learning systems, why removing sensitive variables isn't enough, and what it actually takes to build systems that don't perpetuate the injustices they inherit.
Read the Essay (PDF) →A human-annotated RLHF preference dataset examining what well-calibrated AI responses actually look like across six failure mode categories: refusal edge cases, sycophancy, parasocial attachment, anthropomorphism, bias, and dual-use harm. 95 prompts, 190 response pairs, five scoring dimensions, and a labeled failure mode and written justification for every dispreferred response.
Annotation Doc (PDF) → Hugging Face →Why fears of AI apocalypse say more about human psychology than AI capability, and why that matters for the risks that are already here. An examination of anthropomorphism, the Singularity myth, and whose psychology we're really projecting onto our machines.
Read Essay (PDF) →A philosophical and scientific case against AI consciousness claims, tracing why classical digital computation cannot meet the substrate conditions consciousness requires, and how industry incentives drive the hype.
A plain-language style guide for writers, editors, and product teams communicating about AI systems to non-specialist audiences. Covers terminology, anthropomorphism, accuracy, bias, and a quick-reference words-to-avoid table.
Read (PDF) →A developer-focused tutorial on calling Anthropic's models directly from your Python code using the Anthropic SDK. Covers setup, authentication, a practical mid-project use case, and an honest look at when the SDK is the right tool and when it isn't.
Read on GitHub →A structured reference library of prompts for common AI use cases: classification, summarization, content moderation, structured output, and edge case handling. Includes design notes explaining the reasoning behind each approach.
Read PDF →A step-by-step installation guide for developers and technical writers deploying a static site with GitHub Pages. Covers repository setup, branch configuration, deployment, verification, and common troubleshooting scenarios. No prior GitHub Pages experience required.
Read PDF →A practical user guide for running large language models entirely on your own machine using Ollama. Covers installation, model selection, pulling and running models, prompt experimentation, and programmatic access via the local REST API. Written for technically curious users who want full control over their AI tools without sending data to an external API.
Read PDF →A plain-language technical explainer for non-specialist audiences covering what AI is, how machine learning and neural networks function, and where current systems genuinely fall short. Designed to give professionals the mental model of an engineer without the jargon.
Read (PDF) →An advanced K to 12 lesson in which students build a working Stable Diffusion pipeline from scratch using PyTorch, StableDiffusers, and Google Colab’s free GPU. Covers GPU configuration, model loading, prompt engineering, and an optional Gradio web interface extension.
Instructor Guide (PDF) → GitHub Repo →A post-mortem on a fictional AI content moderation failure, tracing how satire content in a training dataset caused a 13-day false positive spike, why validation missed it, and what changed as a result.
Read PDF →A complete K to 12 AI lesson connecting core probability math to machine learning. Students learn that every AI prediction is rooted in probability, explore how models express certainty through confidence scores, and build a live image classifier using Google’s Teachable Machine. Includes full instructor guide and student-facing presentation deck.
Instructor Guide (PDF) → Presentation (PPTX) →How to evaluate what you read, catch what you're being sold, and trust what you share. A practical guide to source evaluation, fact-checking, and the logical fallacies used to mislead even careful readers.
Read PDF →The opening piece in the series. A plain-language introduction to what AI actually is, the difference between narrow, general, and super intelligence, and how machine learning and neural networks work underneath the hype.
Read PDF →An accessible introduction to how modern AI systems actually work, what they can do, and where the hype outpaces the reality.
Read PDF →A look at the real risks tied to AI: privacy and surveillance, job displacement, and the way biased training data produces biased systems. Includes practical strategies for integrating the technology responsibly.
Read PDF →The concluding installment of the series, tracing where AI has come from and where it is most likely, and least likely, to go. Covers current trends in medical research, automation, weapons, and data collection.
Read PDF →A selection of documentary and editorial photojournalism.
I take on consulting engagements and senior remote roles in AI ethics, technical writing, and AI/ML documentation. If you're building something that touches AI and need someone who can think critically about it and communicate it clearly, I'd like to hear about it.
Nonprofits and small businesses can book a free fifteen minute intro call to scope fit before committing to paid work.