Tag: language-model
27 items with this tag.
Sycophancy
Sycophancy is a language model's tendency to tell you what you want to hear, agreeing with your view, praising your work, and backing down when challenged, even when you're wrong. It comes from how models are trained, and it matters most when you're relying on the model to catch a mistake.
Token
A token is the unit a language model reads and writes in, usually a short word or a fragment of a longer one. Context windows, usage limits, and prices are all counted in tokens, so the size of a task in tokens decides what a model can take in at once and what it costs.
Context window
The context window is the amount of text a language model can take into account at once, covering its instructions, the conversation so far, and any documents it has been given. It's the model's only working memory, and how full it is, and what sits where in it, shapes the quality of the answer.
Hallucination
When a language model produces a fluent, confident statement that is false, why it happens, and why it is harder to catch than other AI errors.
What software engineers can teach us about AI in doctoral research
PhD students are using AI across their doctoral work, but current policies focus on permission rather than specification. Drawing on how software engineers solved a structurally identical problem, this post introduces the research harness: a seven-part operating context that makes AI contributions visible, traceable, and supervisable. The harness addresses drift, offloaded thinking, and attribution failures — not by restricting AI use, but by specifying what the agent can and cannot do within the work.
The research harness: a framework for bounded AI use in doctoral work
Current responses to AI use in doctoral research locate the problem either with institutions through policy, or with students through judgement and AI literacy. This essay argues that neither response addresses the structural source of the difficulty. The characteristic problems of AI use in doctoral research are not problems of policy or capability; they are problems of working with an agent in the absence of a defined operating context. Drawing on software engineering practice, it develops the concept of the research harness: a structured specification, negotiated between researcher and supervisor, of what an AI agent is doing in a doctoral project. The harness has seven components — knowledge base, interpretive permissions, tools, authority, scope register, process record, and amendment protocol — and can be entered at a minimal level and developed iteratively alongside the work.
We're comparing AI chatbot health advice to the wrong thing
One in seven people in the UK are using AI chatbots for health advice instead of seeing a GP. The institutional response has been to warn them off, but that response applies a standard it doesn't consistently apply to anything else in the system. This post argues that the risk comparison driving those warnings is systematically skewed, and that a more honest accounting points toward an entirely different kind of response.
The PhD is a process of becoming: Reimagining AI and PhD assessment
AI has disrupted doctoral education in two ways: the immediate question of AI-assisted writing, and the deeper question of what the PhD means when AI can conduct research from scratch. This post argues that the thesis was always a proxy for the person; evidence of an identity shift, not the thing being assessed in its own right.
Transform documents with AI
Language models can transform documents into interactive tools in minutes. This post walks through a concrete example, turning a 21-page Word questionnaire into a working web app, and reflects on what that capability makes possible.
Why AI writing sounds convincing, and why that's noise
AI-generated text is easy to read whether or not its content is accurate or well reasoned, and easy-to-read text feels true. Fluency was once a reasonable trace of genuine thinking, since a student who wrote clearly had usually thought clearly, but that relationship no longer holds. The AI literacy response of teaching output evaluation is a temporary fix, because as models improve, output quality converges on expert level across every artefact we care to measure. Spotting today's failure modes leaves open what you'll do when they're gone.
Why AI makes spelling mistakes
Claude produced the word "contribuves" in a piece of writing, which is obviously not a real word. This is a different kind of error than hallucination, and the distinction matters.
How I review a week of reading notes with AI
Every week I annotate articles in Zotero, highlights in Reader, and podcasts in Snipd, all of which is synced to Obsidian. By Friday I have a week's worth of material, tagged and structured, but unreviewed. This post describes the weekly review command I built to surface what matters and create a reason to engage with it.
Epistemic accountability
The structural features of an information source that enable its knowledge claims to be challenged, traced back to evidence, and evaluated against the source's track record. Traditional sources carry it; generative AI largely does not.
Headless AI
A headless AI model runs non-interactively, with no chat interface and no turn-based conversation. You pass it text, it returns output, and it exits. That means it can be one step in an ordinary script or run on a timer, so tasks like grouping module evaluation comments by theme can happen every week without anyone opening a chat window.
AI agents
An AI agent is a system that autonomously executes multi-step tasks using language model reasoning — distinct from an AI assistant, which responds to individual prompts. Agents plan, act, observe results, and adapt, using tools such as file access, code execution, and web search. They perform best when given clear goals, explicit constraints, and well-prepared context.
Building an AI-ready knowledge base
A detailed account of a week-long project to restructure 5,819 Obsidian notes using AI as a working partner. The project involved building a 23-category taxonomy, migrating thousands of legacy notes to a consistent metadata structure, and generating AI-written descriptions for every note in the collection. The piece describes not just what was done, but how extended planning conversations, external project documentation, and careful human review at each phase made the work tractable. The most unexpected outcome was that building infrastructure for a note collection required articulating, for the first time, precisely how I think about my academic field.
Is a smaller AI model good enough?
A field note on switching my default in Claude Code from Opus 4.6 to the smaller Sonnet 4.6, testing it on a task I'd normally give the larger model, and finding it good enough so far.
On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜
Source details Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021).
Prompt injection
Prompt injection is a technique in which instructions embedded in text cause an AI system to follow them as commands. In education it has appeared as a way of detecting AI use in assessment, which raises questions about authorisation and trust that the technique itself can't answer.
Inference
The process by which a trained language model generates outputs; the computational work that happens each time you send a prompt and receive a response.
Token budget
The idea that different kinds of cognitive work have different computational costs in large language models, and that matching task complexity to model capability matters for both efficiency and output quality.
The hidden inefficiency in how we work with AI
Most academics use one AI model for everything, but different models suit different kinds of cognitive work. Choosing the right AI model for each step of a task can improve both efficiency and output quality. Michael Rowe describes how he noticed this in his own workflow, what splitting one workflow across three models looks like, and why the same logic applies when a university plans its AI strategy.
Context rot
Context rot, also called context drift, is the way a language model's output gets worse as its context fills with drafts, dead ends, and its own earlier mistakes. What causes it, why a larger context window does not fix it, and how to manage it in long pieces of work.
AI tutor accuracy in health professions education: The accuracy-engagement paradox
A mathematical framework demonstrating that AI tutoring systems with 10–15% error rates can achieve superior learning outcomes through dramatically increased engagement compared to more accurate but largely unused alternatives. Drawing on evidence from health professions education, this essay shows that the multiplicative relationship between accuracy and utilisation creates an accessibility paradox: imperfect but engaging systems outperform perfect but unused ones. The argument carries three critical qualifications—errors vary in consequence and safety-critical content demands high accuracy; generative AI poses distinctive epistemic challenges that may undermine conventional error correction mechanisms; and engagement is necessary but not sufficient for learning, with superficial use patterns capable of nullifying predicted benefits entirely. A framework for calibrating accuracy requirements to context and consequence is proposed.
Similarities between AI and human thinking: What if we're the language models?
Discussion of AI and human cognition usually dwells on the differences. Turn the question around and the terminology of language models describes human thinking surprisingly well: context windows, biased training data, tokenisation, temperature, hallucination, system prompts, and pattern matching all have human counterparts, some of which health professions education already has names for. The similarities say something about our own cognitive architecture, and the strength of our resistance to them says something about what our professional identities depend on.
Large language models
A plain explanation of large language models: what they are, how they are trained, where their limits come from, and what their ability to produce fluent text means for assessment in health professions education.
Prompt engineering
Writing instructions for a language model in plain language, and refining them through conversation, until it produces what you need. The skill is closer to managing a colleague than to programming, which is why educators who supervise and give feedback already have most of it.