Staying the Author of My Own Choices

At our office town hall, we were asked to name a song we loved and how we found it. I mentioned a song I was obsessed with and said I had discovered it through a Spotify recommendation. On the way home, I realised how easily I had called that “discovery”. Spotify had simply put a song it was confident I would love in front of me, and I took it from there. Then I started noticing the same thing everywhere: YouTube choosing what plays next, Instagram resurfacing things I had barely looked at. It made me wonder: how much of what I like is my choice, and how much has been chosen for me?

It turns out there is a name for this feeling: the Algorithmic self. In a 2025 paper in Frontiers in Psychology, researcher Jeena Joseph describes how constant interaction with AI can gradually shape our preferences, self-awareness, and even emotions. Algorithms do not just reflect who we are; they can quietly shape who we become. Each time we accept a suggestion, the system learns a little more about us, and the apps, platforms and predictions around us nudge our sense of self a little further.

I do not think there is anything wrong with being recommended a song I end up loving. What troubles me is how difficult it can be to tell where my choice ends, and the recommendation system’s influence begins. These systems are built to predict what will keep me listening, watching, and scrolling. As Daron Acemoglu argues, AI is exceptionally good at recognising patterns in what has already happened. The problem begins when prediction stops reflecting my choices and starts shaping them.

And yet, this is not the only way I experience AI. At work, I use AI agents, and they have made me think differently about what our relationship with AI could look like.

An agent differs from a recommendation system, which predicts what I might want, and a chatbot, which responds to my questions. An AI agent can take a goal I set, break it into steps, gather and process information, use tools and produce an outcome. It can carry out multiple steps independently, guided by my instructions and the outcome I expect.

At work, I’ve used agents to create reports, presentations and LinkedIn posts, taking each from initial research to a finished output. Working with agents has made me think about three questions that matter when we bring AI into our work and decision-making.

Whose goal is it? A recommendation feed is designed around the platform’s objective of keeping me engaged. An agent works towards a brief I give it. If my brief is vague, so is the result, which forces me to think more clearly about what I actually want.

Can I see how it got there? A feed rarely tells me why it showed me something. An agent can show me the steps it took, making it possible to question its assumptions, spot mistakes and change direction.

Can I say no? With a feed, I can skip or scroll past, but I cannot change what it is trying to do. With an agent, I can stop the task, reject an output or rethink the brief. The AI may do the work, but I still decide whether the result is useful.

This is what working with AI has taught me: its value is not simply in getting things done faster. It is in letting AI handle what it does well while keeping human judgement where it matters. Calvo and colleagues make a similar case, arguing that AI should be designed to support human autonomy (our ability to make our own choices) rather than quietly take it over.

But that autonomy is not guaranteed, you see If I start approving every output without reading it, I am back where I started: accepting whatever is put in front of me. Perhaps the real challenge is not avoiding AI’s influence but staying conscious of it. AI can put possibilities in front of us, but we should still be the ones deciding which ones to take.

I still love that song. I will probably keep listening to whatever Spotify recommends next. But I want to remember that a recommendation is not a decision, and an algorithm is not an author.

I am.