Category: ai

Unexpected AI Detectives: Colouring Book Fans

I’ve been interested in what Charles Goodwin calls “professional vision” for a couple of decades now. How do archaeologists learn how to see and understand archaeology? How do they translate and transmit that understanding to others using various media? It’s intrinsic to so much of archaeological knowledge production, particularly with different recording systems.

It’s become increasingly important to media literacy that we all develop a kind of professional vision of a different sort–detecting AI in the media we consume. For archaeological subjects it is generally extremely easy for archaeologists to identify AI, though this will potentially become more difficult as generative AI improves/consumes more archaeology-specific media.

So thinking critically about strategies wherein we can recognise generative-AI in images is incredibly important, particularly within a discipline that relies so heavily on the visual record. I found some unexpected inspiration on reddit, on the adult colouring subreddit. There are continual queries regarding AI colouring books, asking if images are AI, if it matters, and how to detect it within colouring books. It turns out that colouring helps guide and develop a professional vision–it’s another forcing function, as we identified in our paper on the importance of drawing in archaeology. The act of close attention, of colouring in the particular subject, it becomes obvious when the logic of the image does not hang together.

Take, for example, the image in the header. There’s a lot to annoy both the person colouring the page and the archaeologist. The breaks on the jugs don’t really work, the lines around the temple in the middle aren’t enclosed, so are annoying to colour. Though tbh I always fly around in a hot air balloon with a skull on it, so it got that right at least.

In pursuit of efficiency and digital workflows, many archaeological projects now trace photographs instead of drawing outright. There are some pedagogical problems with this, but it might actually help detect AI images to trace them to see if the various objects within the image actually fit together.

I also like the collective approach (I know, I would, right?) wherein people post on a different subreddit, RealOrAI, who have their own set of strategies to detect generative AI imagery and discuss their opinions on the topic.

Anyway, I could have spent some time trying to refine and improve the above image, and elements are fun (weird) and potentially inspirational. But you do lose out on the potential to connect with archaeological knowledge production in a different way. I don’t think it’s ever a waste to try to do something creative with your research, and drawing is a (scientifically proven) invitation for your brain to see and understand the past in new ways. The folks in the colouring subreddit are serious in their attempts to push the medium and represent different materials in new ways.

And who knows? Coco Wyo might just revive the honoured tradition of isometric drawing in archaeology.

Coco Wyo colouring page with isometric representation of a cozy office with a cat.

(this post was not sponsored, but if you want to talk about a cozy archaeology colouring book Coco Wyo hmu)

Photography & Archaeology in 2025

Happy World Photography day!

I’ve been thinking about digital archaeology and archaeological methods in general as a way to foment intimate connection (no not like that) and prefigure social relations and interpretation. Arguably this began with my work on archaeological photography, inspired by my own practice and taking a class with the brilliant Nancy Van House at UC Berkeley. For my final paper in the class I used methods from Visual Studies to analyse photographs from Çatalhöyük, in a content and semiotic analysis and applied lessons I’d learned to my own practice.

Taking photographs of whole people. Taking photos of people in their power. Asking permission. Working alongside people and earning their trust before taking photos, when I could.

I published the article eight-ish years after I’d originally written it, in Internet Archaeology. There’s more in my thesis, wherein I had an entire chapter written about the many lives of a digital photograph of a coffee pot on an archaeological tell. Learning how to read photographs was an excellent skill, wherein you can get a pretty good idea of the social relations, conditions, and general expertise employed at any archaeological site at a glance.

My photographs have been reusable under a Creative Commons Attribution (CC-BY) license forever, so I occasionally see them pop up as illustrating news stories or other archaeological items of interest. And they’ve probably been scraped for use in generative AI, which was definitely not envisioned when I received permission to take the photographs in the first place. All of the carefully cultivated relationships built with local site participants, my colleagues, students on fieldwork projects, etc have taken on a different gloss now they’re part of a large, decontextualised training set for facial recognition systems.

A search box to see if IBM used your photos as part of a facial recognition program.

After the rapid proliferation of digital photography and rapid acquisition of good quality digital cameras, there seems to be a general decline in quality and availability of archaeological photography. A recent thread on Reddit asked after good quality cameras to take on fieldwork and most people said they just use their phone. There’s certainly further evidence that photography in archaeology continues to proliferate, even while the quality is reduced. I’ve previously argued that one of the affordances of digital cameras, being able to see the photo before you take it, allowed people to take photos with more strategic care, it seems that the lack of training in photography has overcome that minor advantage. Perhaps even further that archaeological photographs are, fundamentally, visual arguments, and the interpretive heft has been overcome by mechanical practice.

Finally, there are relatively few places to broadly share archaeological photographs. Flickr was great until the business model dictated for the company yell at me about retroactive violations of their upload limit. Wikimedia would be an obvious next choice, and I probably should move my collection of photos over there, perhaps with the Flickr2Commons tool. A lot of the most charismatic photos these days are likely to be shared amongst site participants, in whatsapp groups, and that’s probably fine. But between AI facial recognition tools and the lack of hosting, we’re in danger of becoming an un-faced discipline again, finds and sites divorced from the people who participated in their uncovery.

Finally, my own practice has faltered in recent years, as the demands build up. Digital archaeology is laughably difficult to personalise, with many variations of people pointing at screens, like this photo from the 2024 Heritage Jam:

A group of people clustered around a screen, smiling.

Indeed, much of it is difficult to disambiguate from generic marketing imagery.

Of course, I’ve taken hundreds of these kinds of photos as well, so perhaps it’s me that needs a more creative approach:

An archaeologist recording at Hili-16 archaeological site in the UAE. There are some archaeologists in the background and it's a sunny day.

So what’s next for archaeological photography? Are we retreating to our private groups and stashes, with only the bare (faceless) minimum published, uploaded to archives and shown at conferences? Does the preponderance of generative AI make the above photographs obsolete? Do we have enough photographs of people pointing at screens and holes, now that the imagined truth of photographs has been completely undermined?

Smile! click

Generative AI and Archaeological Visualisation

I wrote a brief article for The Conversation on the challenge that AI presents to archaeological visualisation. Check it out here:

https://theconversation.com/how-ai-imagery-could-be-used-to-develop-fake-archaeology-247838

Though I have done a lot of scicomm, it was the first time I’d written for The Conversation. At the outset, my goals for the article were:

  1. Point out that generative AI was being used increasingly to illustrate the past
  2. Note that many examples of generative AI seem to be based on faulty and misleading images that have been heavily critiqued by (primarily feminist) archaeologists
  3. Argue that illustration is actually an amazing tool to understand the past, and it’s a shame to delegate it
  4. Propose that there are creative ways to engage with generative AI in archaeology, and these may become more feasible and ethical through lower impact AI. As Haraway notes, I would rather be a cyborg than a goddess.

I’m not sure that it properly covered all of these goals in the end, but it was an honest try. The article took a bit of a left turn in tackling the growing problems of representation in archaeology, race science and pseudoarchaeology. Additionally some of the original references fell away in the edit, and some sentences were added in that I didn’t properly check before it went to publication.

Working with popular media is always going to be a bit of a gamble and most of us could use more practice. Working with The Conversation was a considered move though, as it seemed like a subject that would be of broader interest than my little blog. So far, so good. Perhaps I’ll do better in the future.

Indeed I firmly believe I’ll be able to write a potentially completely different follow-up to the article in a year’s time, after we teach the first iteration of our new course, AI and Archaeology as part of our Digital Heritage and Digital Archaeology MScs. I’m also excited for the MAIA project to get going, so there will be a lot of energy in this particular subject.

Thank you to Kathryn Killackey for reading a draft of the article and suggesting changes, all mistakes are my own.

Hire Katy for cool illustrations based on science:

https://www.killackeyillustration.com/contact

The (boring) alternate realities of ChatGPT

Inspired by a shocking story of a list of fake references provided from a senior to a junior scholar, (details redacted obviously) I thought I’d ask ChatGPT to generate 10 references for me. Interestingly, one out of the ten is real (2), the others are not but potentially plausible. I guess, except for me co-authoring with Colin Renfrew:

  1. Morgan, C. (2012). Digging digital: Reflections on twenty-first century archaeology. In Clare Mills, Michael Pidd, and Esther Ward (Eds.), Proceedings of the 2012 Digital Humanities Congress (pp. 1-11). Sheffield, UK: HRI Online Publications.
  2. Morgan, C., & Eve, S. (2012). DIY and digital archaeology: What are you doing to participate? World Archaeology, 44(4), 521-537.
  3. Morgan, C., & Wright, H. (Eds.). (2018). Archaeology and digital communication: Towards strategies of public engagement. London, UK: Taylor & Francis.
  4. Morgan, C., & Eve, S. (Eds.). (2017). Digital archaeology and the archaeological network. London, UK: Taylor & Francis.
  5. Morgan, C., & Hacıgüzeller, P. (Eds.). (2020). Archaeological futures: Methods, data & practices. London, UK: UCL Press.
  6. Morgan, C., & Kamash, Z. (Eds.). (2020). Material cultures of remembrance: Exploring the archaeology of memory. London, UK: Routledge.
  7. Morgan, C., & Renfrew, C. (Eds.). (2019). Digital dawn: The dawn of digital archaeological practice in the Middle East and Mediterranean. London, UK: Oxbow Books.
  8. Morgan, C., & Ross, C. (2018). Digital media and archaeological objects: A challenge for interpretation. In Erin Walcek Averett, Jody Michael Gordon, and Derek B. Counts (Eds.), Mobilizing the Past for a Digital Future: The Potential of Digital Archaeology (pp. 97-118). Grand Forks, ND: The Digital Press at the University of North Dakota.
  9. Morgan, C., & Smith, C. (2016). Emancipatory digital archaeology. Antiquity, 90(354), 713-718.
  10. Morgan, C., & Winters, J. (Eds.). (2019). Teaching Digital Archaeology: Methods and Practices. London, UK: UCL Press.

Most of the papers sound pretty banal, except for maybe the one authored with the incomparable Zena Kamash on memory and futures. Some of them represent things I probably should have done, like publish part of my thesis in Antiquity (9), though it looks like I’d have to co-author with Claire Smith to do so.

Perhaps I could have engineered a better prompt…but maybe I’ll just go and write instead.