Human-centred intelligent systems postgraduate personal statement example

Personal statement example

Two years of answering calls about water bills has taught me more about intelligent systems than any lecture did. Our department uses an automated tool that flags accounts as likely to fall into arrears, and advisers are expected to act on the flag. It is usually reasonable. But I could not explain to a caller why she had been contacted, only that the system had suggested it, and she quite fairly wanted to know what had been assumed about her. That gap between a model that performs well and a system a person can understand and argue with is what I want to study properly.

My degree in Computer Science gave me the technical grounding: machine learning, databases, software engineering, and a human-computer interaction module that I chose largely out of curiosity and ended up caring about most. For my final-year project I built a voice-controlled kitchen timer on a Raspberry Pi, using an off-the-shelf speech recognition library rather than anything I trained myself. The interesting work was not the build. It was watching eight volunteers, including two people in their seventies, try to use it. They spoke in full sentences when the system wanted keywords. They paused mid-phrase and were cut off. One participant repeated herself more loudly each time, which made recognition worse, not better. I had assumed error rates were the measure of quality; what actually mattered was whether the system gave any signal about what it had heard, so the user could correct course. Adding a simple spoken confirmation of the parsed command reduced task abandonment noticeably in my small sample, though with eight participants I was careful not to claim more than a trend. I learned as much about the limits of my own study design as about voice interaction.

Since graduating I have kept reading rather than letting the subject go. Don Norman's The Design of Everyday Things gave me vocabulary for what I had observed about feedback and mental models, and it holds up surprisingly well when applied to conversational systems that offer almost no affordances at all. I have also worked through material on explainability, including the distinction between explanations that faithfully describe a model's reasoning and those that are merely plausible to a human reader. The second kind troubles me: a comforting explanation that does not correspond to how a decision was actually reached seems worse than none, because it discourages challenge.

My volunteering has sharpened this. At a Saturday drop-in at my local library I help people with phones, forms and online accounts. Most difficulties are not about literacy or ability; they come from systems that assume a confident user with time, a stable address and no interruptions. Watching someone navigate a chatbot that keeps offering irrelevant options is a useful corrective to designing for people like myself.

I am applying for a master's because the questions I have reached are research questions, and I lack the methods to answer them. I want structured training in evaluation design, qualitative as well as quantitative, so that my next study is not eight friends in a kitchen. I would like to develop work on how speech and decision-support systems communicate uncertainty to users who have no technical background and often no real choice about using them. In the longer term I hope to work in design research within public services or utilities, where automated decisions reach people who did not opt in. My call-centre years are not a detour from that ambition; they are the reason I have it, and they mean I will bring real users' voices into technical discussions rather than imagining them.