Shifting Realities

Shifting Realities explored what happens when artificial intelligence moves beyond being a technical tool and begins shaping how people are observed, understood and influenced. Rather than answering that through one medium, I built the project as three connected investigations:

1. In the Public: making surveillance something you experience

The first project began with Foucault's Panopticon: surveillance changes behaviour not because someone is necessarily watching at every moment, but because people know that they could be.

I translated that idea into Under the Gaze, a cyberpunk visual story following Cam through a city where facial recognition, location tracking, behavioural profiling and targeted systems are embedded into ordinary life. The project asks what happens when the observer is no longer a guard in a tower but an invisible network of automated systems.

The visuals were generated with MidJourney and developed through iterative prompting, selection and post-production. That process became part of the project itself: AI made rapid image production possible but struggled with character consistency, spatial control and sequential coherence. I had to repeatedly generate, curate and edit outputs to construct a narrative that actually held together.

I didn't want surveillance to remain something happening only to the fictional character, I therefore built an interactive consent flow in HTML, CSS and JavaScript, deliberately borrowing the formality and friction of real digital terms-and-conditions interfaces. It asked users for personal information and consent before participating in an emotion-tracking experiment, making the act of data collection visible rather than hiding it behind the experience.

I then experimented with facial-emotion analysis using OpenCV and DeepFace, processed the resulting data in Python and visualised it using Matplotlib. The interesting result was how often its classifications failed to match what participants said they actually felt. That failure strengthened the conceptual argument: turning human behaviour into data does not automatically mean understanding it.

2. The Awakening of Ava: intelligence isn't the same as consciousness

Using Ex Machina as a case study, I wrote a critical media analysis around the distinction between performing intelligence and possessing consciousness. Ava appears to understand emotion, manipulate relationships, pursue autonomy and act according to her own interests. I used that fictional scenario to investigate several questions that become increasingly important as AI systems become more socially convincing:

Can fluent behaviour be mistaken for understanding?

What happens when people form emotional attachments to machines?

If an artificial system ever became genuinely autonomous, where would ownership end and rights begin?

And how should accountability work when an intelligent system acts beyond its creator's intentions?

The value of the exercise wasn't predicting whether Ava-like AI will exist. It was using fiction to expose assumptions that already affect how people interpret real AI systems: we are extremely willing to infer understanding, intention and personality from convincing behaviour.

3. Will Robots Finally Revolutionise Education?: moving from speculation to a social system

The final investigation moved from fictional AI into a real institutional context: education. I explored the argument that increasingly capable AI and robotics could support personalised learning, reduce administrative work and extend educational access - but also introduce new problems around privacy, bias, dependency and control.

Rather than finishing with my own argument, I collected anonymous responses to deliberately expose it to other perspectives. One participant argued that automation might be most valuable precisely where teachers are least human, while preserving the relational parts of teaching that people remember. Another raised a more difficult possibility: hyper-personalised education could optimise around each child’s measured strengths and weaknesses, but in doing so could also begin deciding very early what sort of future each student is suited for.

At first glance these are three different questions, the thread connecting them is agency.

Who gets to observe us?

Who gets to interpret what our behaviour means?

When does a machine's simulation of understanding become persuasive enough that people treat it as real?

And what happens when those systems move from describing people to making decisions that shape what happens to them?

Shifting Realities was an early project, and some of its methods were exploratory rather than experimentally rigorous. But it established a question that has carried into much of my later work: When a system makes claims about human behaviour, what gives us reason to trust those claims, and what happens when we do?