We're building something 'irresponsible'! – Part 2

The detour 

The students' first idea was facial recognition: smart glasses that scan a face and pull up personal information on the spot. They got far enough to see how it might work, then ran into a wall that was both technical and ethical. The project's whole premise was to make people uncomfortable without actually harming anyone, and real facial recognition broke that rule. So in sprint three, they dropped the idea and went looking for a different way to make the same point.

A trial run at SURF

What they landed on was a fictitious role-play game. Each participant wears a QR code carrying a name and a backstory. One participant secretly uses smart glasses to scan codes and pull up profile data, scores, and reviews written by other "users." The group then has to pick a class representative, a role with real influence over the rest of the group, based on nothing more than a short introduction and first impressions. Everyone but the glasses-wearer is making that call blind.

They ran an early version of this with SURF colleagues as test participants. The feedback from that session reshaped the role-playing a lot! 

The pivot

The SURF test showed the team where more tension in the role playing rose. So they made three changes: an imposter, one participant who lies about their own background to get picked, raised the stakes of trusting first impressions. Shifting from visual profile-reading to audio, heard privately in the glasses-wearer's ear, made the information feel more invasive and harder for the group to challenge. And dropping the teacher role kept the power imbalance strictly peer-to-peer.

The Dark Tech Expo

At the Dark Tech Expo at the HvA, the students got to present and test their refined scenario with their peers. Colleagues came along too, and we recorded a short interview with the team. Our colleagues joined us to see more about the Dark Tech Studio's work and learn what experiments were being done. 

We also got to interview one of the students for their perspective on how the project was going and what they had learned regarding smart glasses and responsible technologies. 

You can watch that interview here: Link 

With further refinement in place, scenario participants vote now on a top-three class representative after a round of introductions. The whole time, the glasses-wearer is privately fed scores and reviews through QR scanning and audio, deciding who to push for based on information nobody else in the room can hear or question. Some of that information may be a bit incorrect. Nobody finds out until after the vote!

The Handover

In June, the student team delivered their final advisory report and handed the prototype over to us.

Two findings stood out. The students expected the camera to be the unsettling part, but it wasn't. The audio channel felt far more invasive: information delivered privately, in a strange voice, with no one else around able to hear it. They also saw automation bias play out directly. Participants trusted glasses-fed information and acted on it even when it was fabricated, simply because it arrived through a system that looked confident and "data-driven."

The students' advisory report, based on all of this, is now public and available for anyone in our communities who wants to dig deeper: Link

Building on Corrupted Visions 

With the students' work as inspiration, our Responsible Tech team is now taking the next phase forward ourselves.

We're designing scenarios around moral dilemmas that people in education and research are likely to run into in their own institutions. We're using both Meta Ray-Ban and Even Realities glasses to show our audiences various capabilities they should be aware of.

One example: we now have scenarios built around the Even Realities G2, which has an integrated display. In one, the wearer is in a job interview; in the other, a social conversation between friends. In both, the glasses feed AI-generated answers and coaching tips directly to the wearer through the lens, live, as the conversation happens.*

* Technical note: the AI model is Google DeepMind's open-weight Gemma4 E2B model, running completely locally on our laptops. It is fed audio from the conversation and outputs the answers. No data leaves the workshop, and nothing is stored.

These scenarios surface a few ethical tensions that only really become obvious once you're inside them:

• Information asymmetry between wearer and non-wearer
• AI-mediated social performance (not unlike colleagues writing AI-mediated emails)
• Erosion of authenticity and autonomy

We're experimenting with what is now possible with various glasses technologies. This isn't a hypothetical future, it's already on the market and on people's faces. We do this so our members can try it themselves, then discuss and reflect on what it means for their work and communities in relation to these devices.

First test at Media & Learning 2026

Last week, we ran the first public version of this format at Media & Learning 2026 in Leuven, our first time running the workshop with an audience that wasn't our own colleagues.

After a presentation on ethics, public values, and smart glasses, our team helped fifteen educators and IT specialists work through the scenarios in pairs or small groups. Each group had one pair of glasses: two scenarios used Meta Ray-Ban glasses (no display), and two used Even Realities G2s with waveguide lenses.

Groups rotated through the scenarios, experiencing sound, on-lens text, and QR scanning across different role-plays. As with the student version, we designed these to feel concerning or emotionally charged, so participants would engage with the ethics rather than just the novelty.

We had a full house and a lot of positive early feedback. By a show of hands, this was the first time most people in the room had tried smart glasses.

What stood out most was how the glasses disrupted presence. Wearers had to split their attention across multiple information streams, more so than with a smartphone, and conversation partners noticed the gap. Wearers also felt more confident arguing their case than the person across from them, exactly the power imbalance the students flagged in their advisory report. That asymmetry also made for a quiet sense of mistrust between participants once they realised what the other side could see.

Over summer, we'll keep refining these scenarios, test new ones based on what we've learned, and bring the format to a few more events. We'll pick this up again in part 3 later in the year.

Wishing you a great summer, 

-- The Responsible Tech team, Public Values Program at SURF and the Digital Realities Program