AI Gamification: With Paul against Caesar through the Roman Empire

On my blog, I regularly share (for example here) very critical thoughts about AI development - or rather, how we as a society deal with it ... or, rather, how we avoid engaging with the challenges at all. These concerns somewhat take away the joy of playfully exploring what the technology is already capable of today. Nevertheless, I would like to share an impressive example of a very practical application today. As a university lecturer, I am naturally very interested in pedagogical concepts, including the much-invoked (with its advantages and disadvantages) "gamification" of the learning process. Everyone who teaches others has certainly thought about how to make their teaching more interesting. However, until now, few probably considered developing a computer game for their pupils, students, or other learners as a possibility, as this required considerable specialized knowledge - until now. Newer large language models are actually remarkably good at developing interactive web games.

Ethan Mollick specifically gave me the idea through a post to have an LLM create a game based on one of my scientific papers. I work extensively on the question of how early Christians navigated the dangerous situation in the Roman Empire (see for example here; my most recent book in English on the topic is this one). I have just completed a new paper on how telling subversive counternarratives served to subtly challenge the dominant narratives in Greco-Roman society. (The article hasn't been published yet, you can read about my general thoughts on stories in the New Testament here.) I submitted this manuscript to Anthropic's Claude 3.7 with a very simple prompt ("Create a fun interactive computer game on the basis of the following paper"). The result, which was delivered to me immediately, impressed me greatly. You can easily access the artifact here and play it. (The whole thing is also available - after another prompt - in German, albeit somewhat awkwardly translated, here.) This is what the interface looks like:

Of course, one must add: This works so well because the individual necessary code components are essentially available 1:1 in the training material. It is certainly not the case that all game developers are now suddenly facing unemployment.

At the same time, one should not downplay the significance of such a one-prompt game. We must not forget what programming such a game would have looked like until recently. Ethan Mollick has provided a nice comparison here] While ChatGPT back then still had considerable problems implementing the ship catalog from Homer's Iliad 2.494-759 as an interactive map, the AI agent Manus today succeeds immediately. Accordingly, one must also clearly state that until recently, a game like the one presented here simply could not have been created without proper prior knowledge, even with the help of a large language model. We don't even see yet the actual economic consequences of the potential that large language models already bring. The list of websites of comparable quality that are still commissioned today for thousands of euros daily is enormous. (Not to mention the hundreds of thousands that the German federal government recently put into the development of games - such as this one or that one.)

Above all, in my eyes, attention should be paid to how creatively my paper has been developed further here. The basic idea behind the implementation as a game - that is, the dynamic that on the one hand you gain points the more 'influence' you get in the community, but on the other hand you also receive minus points when you draw unwanted attention to yourself - is simply great. This is exactly the kind of dilemma that a person like the Apostle Paul would have been in. And I think that a great added value for students lies in being sensitized to this problem (which does not affect us at all today in our comfortable situation!) through such a game.

In addition, there is basically the potential to convey historical facts and connections in this way. Of course, one would still need to make some improvements, because some of the scenarios quickly designed by Claude are historically quite inaccurate (the Roman triumph procession, for example, was only celebrated in Rome!). But here too, one should beware of academic arrogance. Because on the one hand, I find it quite impressive that the game calculates an action with different risks depending on the locality (due to differently sized Roman presence) in a historically sensitive way. This is a quite successful implementation of an argument from my paper that has been given too little consideration in research so far.

And finally, the historical distortions also contain a quite salutary challenge to the assumptions with which we as historians work. In our work, we often all too readily forget that we apply anachronistic categories to our source texts, which might often not have been comprehensible to people in antiquity. So it is true that in the game, some religious ideas and some political events from the Greco-Roman environment are 'wrongly' conceptualized. On the other hand, it is not so unrealistic that someone like Paul - who would have analyzed his environment from up close, but with a correspondingly limited perspective - would have entered into interaction with Roman imperial ideology partly with a quite similarly distorted perception.