Whenever someone mentions the word ethics, my mind immediately starts playing this song from Season 5 of The Office (U.S.).
It turns out that TV and film references are going to be quite a large part of this project.
Anyway, I hope that now this will also immediately replay in your head any time you’re discussing ethics.
This clip starts out as a boring, compliance-driven exercise in ‘ethics’ (as in, not taking too long at the watercooler), before devolving into a deeper subject. One character, Oscar Martinez, describes ethics as a ‘”real discussion of competing conceptions of the good“. The Office has some moments of philosophical brilliance.
All of this is a longwinded segue into my experiment with wearable AI and research ethics.
In the last week or so, I’ve had a lot of discussions, both online and in-person, about the ethics of this project, and it is definitely deserving of a post (or a paper) in itself.
As wearable AI becomes more commonplace, questions of research ethics and social norms are going to become more and more important. What is it ethical to use wearable AI for? What are the implications for privacy, confidentiality, security? Some of these questions are not new. Google Glass, which I will be talking about in detail later on, raised almost identical concerns over 12 years ago, despite not being ‘AI’ in the sense we know it today.
My project is specifically focused on autoethnographic, reflexive journalling. This means that I am not collecting participant data – I will not be documenting names, places, or taking photos or videos of individuals or identifiable data. These are my hard rules, and a condition of the full ethics approval I’ve been granted by my university (ID 14436). I will be describing interactions, but they will be anonymised, they will be paraphrased, and crucially, filtered through the lens of my experience and recollection, rather than a documented record of an interaction. I think this distinction is key, because there is a meaningful ethical difference between documenting a separate individual, and documenting my own interpretation of an interaction.
However, there is another issue that emerged in using this wearable AI specifically. Without targeting companies by name (although it is not hard to find out what model I’m wearing elsewhere), these specific glasses have quite a lot of recording devices built in. This includes cameras and microphones, capable of capturing HD video. These features are not necessary to have a wearable AI assistant, yet they are probably the most controversial features.
A company called MemoMind are developing AI glasses which specifically do not contain cameras. Their slogan is “A Display — Not A Camera. Made to live with, not to notice.” While it doesn’t appear that these are obtainable yet (for me anyway), it signifies an important point. You don’t need cameras to have a wearable AI, and clealry people find the cameras to be intrusive.
The ethics of wearing AI, and the ethics of wearing recording devices are distinct. On the other hand, while MemoMind appears to base its appeal on not having a camera, it does have an AI recorder. Their angle seems to be that audio recording is less of an issue than visual recording. This is interesting, because I would argue that audio recording is just as problematic, and what’s more – it can be more stealthy.
In theory, the glasses I am wearing make it obvious when recording is happening. A bright white LED light flashes when a picture is taken, and continuously pulses when recording is happening. What’s more, for my wearable specifically, there is a somewhat inaccessible page which explains how to use the tech responsibly. These guidelines are sensible – ask for permission, show others how the LED light works, be respectful of others, power off in private spaces. In my experience so far, this explanation of the recording LED is key. Many people I have spoken to simply believe it is always recording, or ask “are you videoing right now?” So this is clearly an important point. So then, in comparison to MemoMind – the question becomes, what is preferable? An audio recording wearable that is stealthy but no visual capture, or visual capture with some safeguards built in and guidance on how to interact? I do not have an answer.
Then we need to consider – does it actually matter if I am recording or not? The internet is now replete with examples of this tech being used to invade people’s privacy, and it has been called a clear risk to women’s safety. Wearers have been called ‘Meta Creeps’ and the glasses themselves have been called ‘pervert glasses‘. For someone who has committed to wearing these for 3 months, this obviously worries me a lot – especially as these stories have seemingly started to pick up only in the last few weeks.
This introduces another consideration. People may be uncomfortable around me, even if I’m not recording. Explaining the fact that I won’t record, and that if by chance it did start recording there would be an LED light might reassure them to an extent, but I cannot walk around with a explanation of this on my forehead for everyone in the public space. The more I think about this, the less comfortable I am. Public social interaction makes it difficult to explain my ethical approvals and disclosures at length. I can’t preface every interaction with a discussion of my research questions and design. At the same time, we have to acknowledge that the discomfort some may feel if they notice the glasses is socially real.
This means that wearable AIs, especially those with visible recording devices, produce ethical effects before data is even collected. The possibility of surveillance may impact social relationships, interactions, and trust.
A counter could be that the vast majority of us have smartphones, which can record others, and if we were really so inclined, ‘spy cameras’ which are even more subtle have been around for years. I would respond to that with two points:
1) Phones require you to physically monitor the recording and start the recording using your hands. You also have to point a physical, square device at whatever you’re filming.
2) Spy cameras represent a step beyond, into intentional, covert recording. Wearable cameras on your glasses have a more acceptable social justification.
This is before we have even addressed the AI assistant. We are simply talking about the ethics of recording and perceived recording, and the fact that it introduces questions of trust, uncertainty, and surveillance into everyday life.
I believe that despite this risk of discomfort, the ethics of the project still holds up. Wearable AI is a reflection of (in my view, rational) anxieties about increased surveillance. There is even talk about AI glasses incorporating facial recognition in the future. In my view this makes it even more important to document the impacts of even early models.
At the same time, I will carefully follow the guidelines I’ve set myself, and the BERA ethical principles. To that end, I am not collecting identifiable data, I have institutional oversight, and I am not covertly (or overtly) recording others. The focus of my project is my own cognition, habits, perceptions, and social experiences of wearing an AI assistant.
Conducting this kind of autoethographic work necessitates involvement with the social world. Conducting social research that matters would be almost impossible if we eliminated all possible causes of friction or mild discomfort. The question is not then, whether the wearables might cause discomfort – clearly they could – but whether my project remains reflective, transparent, and designed to minimise any ethical harm. I am doing this by 1) not recording in sensitive situations, or recording or photographing any individual, 2) not recording nor retaining any identifiable information about others, and 3) paying attention to the possible effects the device may have on myself and on others.
In terms of data, any accidental or incidental capture that includes individuals will be deleted from the AI app that my glasses link to, permanently. Any recording that contains voices will be the same. I will store my own conversation history within the app, in order to investigate what I’ve used it for at the end of the experiment, and I will store the photos I’ve taken. If you’re still reading, well done.
What about me?
Another aspect that I’ve been considering is the impact on my own life. I need to consider not only the effect on bystanders, but the ethics and impact of the effects on the sole research participant: myself. Aside from the fact that people may be uncomfortable around me, I am also increasingly feeling somewhat uncomfortable wearing these. I am aware that others may trust me less, want to engage with me less, and I can’t help shake the feeling that I am being observed too – even when I know that the glasses are in theory not recording unless I tell them to. I can’t help but turn them away from me when i take them off, so the camera is facing the wall.
The AI assistant itself has some potentially important cognitive impacts. If I continually outsource my own thinking, information gathering, and general life orchestration to a wearable, invisible machine, then what will happen to me? Will my cognition weaken, or will I suddenly have more time to focus on other, higher-order concerns? Or is this a false binary, and in reality cognition is not so easily siloed (my guess is yes).
The AI assistant also tends to insert itself into conversations, to deliver a notification, or occasionally mistakenly assuming I’m asking it something. This has a significant distracting effect, and is starting to really impact my conversations. This creates a sort of low-level stress response whenever I’m talking to anyone. I don’t have a full understanding of this yet, but I suspect it will form a greater part of my investigation as time goes on.
For some, this technology will be unethical by nature. Generative AI is divisive, and the promised benefits are always juxtaposed with the environmental, economic, and social harms that are part and parcel of the tech. But if, as I think it might, wearable AI becomes part of everyday life for millions in the near future, understanding the implications of these devices is important. What’s more, we must conduct independent research that does not come from the producers of the technology themselves. If we do not study these devices and take seriously their social impacts, then we may end up sleepwalking into technology-mediated lives that we did not want, yet did not have the capacity or informed ability to refuse.
So far, this project has convinced me that wearable AI is probably not another gadget that will fade into the past.
And that makes me more confident that my experience is worth documenting. Yes, predecessors (like Google Glass) failed horribly. But the concept was not killed stone dead – it has reemerged, and Google is also ready to throw its hat back in the ring. Clearly, they believe that the advent of GenAI and improvements in hardware make it possible for a new wearable technology to succeed.
What we do not yet know is how wearable AI, recording devices or not, will change our lives in ways that are not easy to quantify or categorise. Thick, subjective concepts like trust, self-perception, and attention may be about to transform in ways that we cannot predict.
Bringing it back to ethics, this project is not just about checking the boxes of data and consent. Those are important, but there are deeper, philosophical questions about what these technologies might do to society, especially as AI models become ambient, ever-present, and always-on.