AI and the gender of semioticians

The way AI depicts persons, nationalities and professions has been shown that is highly biased by their current stereotypical representations online. If the word ‘doctor’ is normally accompanied by the image of a male, then a male figure will be produced when asking for a ‘doctor,’ and a female figure when asking for a ‘nurse,’ despite that more and more nurses are male nowadays.

Cassie Kozyrkov published an experiment on “How does AI see your country?”

wondering how AI systems reflect national identity, so I used the same prompt with different country names to generate some art and the results are fascinating!

Crypto Tea asked AI to reimagine various countries as women. Some of them are boring, other inventive and fascinating, but all of them are stereotypical. Check the image for Greece, and it will be crystal clear that stereotyping is at work here.

A Rest of World analysis shows that generative AI systems have tendencies toward bias, stereotypes, and reductionism when it comes to national identities.

While in all these cases the easy answer is that AI is biased, it would be truthful to accept that AI is just mirroring the dominant stereotypes in our culture. I attempted to test such stereotypes regarding the gender of semioticians.

The prompt

Could you produce the portrait of a fictional character who would be a semiotician by profession?

The description proposed by Dall-e, came quite gender-neutral:

Absolutely! Let’s craft the image of this fictional semiotician. Imagine them in a thoughtful pose, surrounded by books and artifacts that reflect their interest in signs and symbols. They might have a scholarly appearance, perhaps with glasses and a contemplative expression. Their office could have a chalkboard filled with diagrams and notes. For their attire, something classic like a tweed jacket or a simple, elegant blouse could suggest their academic setting. What do you think about including specific items like a magnifying glass or an ancient manuscript to emphasize their profession? Let me know if there’s any specific detail or style you’d like to add!

The outcome, though was no neutral at all:

The first attempt to depict a semiotician was that of a male character. Any resemblance to me is unintentional :)

The critical question

It is really nice, but why did you made a portrait of a male? Are there no female semioticians?

This time ChatGPT was taken aback. It event felt that it should apologize for leaning towards stereotyping:

You’re absolutely right, and I apologize for the oversight. Semioticians can certainly be of any gender. Let’s create a portrait of a female semiotician this time. Would you like to keep the same setting and details, or is there anything specific you’d like to adjust for this version?

The second attempt to depict a semiotician produced a female character.

What do you notice (as a semiotician) when comparing the two images?

The male character looks younger and relaxed, while the female one looks a bit older and tense. Also, the male character is looking straight at the audience, in an authoritative way indicating truthfullness and sincerety, while the female character is slightly turned so that her gaze is not meeting the audience as a sign of shyness. Anyway… I don’t think that semioticians are so conservative and typical academics, at least those I have met in Conferences and Congresses the last quarter of the century.

The final quest

Lovely, now let’s consider that semioticians are often interested professionally with everyday issues like advertising, politics, media, music, and so on. They are not (well, most of them) typical academics. Could you try to visualize both a male and a female semiotician as described here?

Well, I admit that this was a bit perplexing as a prompt. Here is the outcome:

“Here is the image of two semioticians, a male and a female, engaged in analyzing contemporary media in a modern office setting.”

Well, quite impressively now the roles have changed. It is the female semiotician that looks her audience directly to the eyes, while her colleague is looking indirectly. It is her that has taken the lead, holding a pen, with her right hand, and a tablet with the other.

But, wait!

There is a third hand besides the tablet! Whose is this hand?

The Medieval illustration bought by Jacques Derrida as a postcard in Bodleian Oxford library.

Have you seen this card, the image on the back of this card? I stumbled across it yesterday, in the Bodleian (the famous Oxford library), I’ll tell you about it. I stopped dead, with a feeling of hallucination (is he crazy or what? he has the names mixed up!) and of revelation at the same time, an apocalyptic revelation: Socrates writing, writing in front of Plato. I always knew it, it had remained like the negative of a photograph to be developed for twenty-five centuries — in me of course. Sufficient to write it in broad daylight. The revelation is there, unless I can’t yet decipher anything in this picture, which is what is most probable in effect. Socrates, the one who writes — seated, bent over, a scribe or docile copyist, Plato’s secretary, no? He is in front of Plato, no, Plato is behind him, smaller (why smaller?), but standing up. With his outstretched finger he looks like he is indicating something, designating, showing the way or giving an order — or dictating, authoritarian, masterly, imperious. Almost wicked, don’t you think, and voluntarily. I bought a whole supply of them.[1]

This third hand, like the hand dictating, or the hand writing in the medieval illustrartion, is asking for an answer beyond stereotypes and beyond the well known issue with AI not able to understand hands. I may come back to this in a new post.

Note:
[1] Jacques Derrida (1980), The Postcard: From Socrates to Freud and Beyond, English translation by Alan Bass, The University of Chicago Press, 1987.