Challenge on four case studies

The eighth stage of DisclAImer. Final warnings before the revolution took place on 3 December at the University of Palermo, with a focus on the encounter between artificial intelligence and the humanities. Promoted by Corriere della Sera in collaboration with Cineca and Fondazione Mondo Digitale as knowledge partner, the event confirmed the role of the University of Palermo as a national reference point for Digital Humanities, thanks in part to the activities of the Artificial Intelligence & Robotics for Humanities Lab (AIRH) [see the news item Culture meets robots in Palermo].

During the debate, Vera Gheno and Paolo Giordano discussed the relationship between language, creativity and AI: for Gheno, the main challenge is learning how to ask effective questions to generative models; Giordano, on the other hand, warned against the “oracle of the algorithm” and the excessive delegation of cognitive and creative responsibilities. Researcher Arianna Pipitone recalled the famous experiment in 2021, when the Pepper robot made its thought process audible, opening up a new perspective on the “inner life” of machines and the ancillary role of AI in the work of humanists.

The challenge: four case studies to experiment with different approaches to AI-assisted knowledge

In the afternoon, the challenge “Humanities + AI: Unlocking New Possibilities in Culture, Creativity and Thought”, hosted in the Sala delle Capriate of the Complesso Monumentale dello Steri, invited 60 university students to tackle four challenges proposed by lecturers and experts working in the field of Digital Humanities. Below are the four case studies with guidance from the lecturers.

1. Reverse prompting with Salvatore Di Dio, associate professor (architecture)
2. The challenge served to make them aware that communication with artificial intelligence requires precise linguistic and technical skills, far beyond a simple narrative description of what one wishes to achieve. Students were asked to carefully observe seven reference images and then attempt to reconstruct them through textual prompts, discovering the difficulties in translating a vision into accurate instructions that included not only visible elements, but also technical parameters, cultural references and final objectives. The groups were thus able to directly experience the difference between a generic prompt and a professionally structured one, realising the need to refine their prompt engineering skills through systematic training and the study of communication techniques with generative models.
3. Modelling historical evolution with AI methodologies with Salvatore Miccichè, full professor (Physics and Chemistry)
4. Students and doctoral candidates from scientific fields (Physics, Computer Science, Statistics and Data Science) applied their skills and knowledge in Machine Learning, Deep Learning and Natural Language Processing methodologies to design applications capable of analysing texts and open sources (newspapers, social networks, etc.) relating to specific complex legal, social and socio-economic systems. The aim was to extract useful information to characterise their evolution over time and contextualise them in their respective historical periods, putting into practice interdisciplinary approaches to understanding contemporary phenomena.
5. Making AI flourish: iconographic exercises and automatic image generation with Diego Mantoan, Associate Professor (Humanities)
6. The challenge on iconographic exercises served to make students aware that automatic image generation is actually rather limited and basic, unless refined by human curation. The datasets from which artificial intelligence draws are often small and stereotypical, and therefore prone to logical or historical errors when it comes to creating images that resemble works of art. Students were invited to study the works in the University's art gallery and then test different software in an attempt to create new works in perfect 18th-century style, credible from a stylistic, historical and even conservation point of view. The groups were thus able to experiment creatively with the production of new images, realising the need to train and refine generative AI through the study of historical and artistic images.
7. Specialised languages, AI and neural machine translation with Floriana Di Gesù, Associate Professor (Humanities)
8. The challenge on specialised languages was introduced by an overview of sector-specific languages (medicine, law, economics, science, tourism, advertising, etc.), highlighting the horizontal dimension of the sectors and the vertical dimension of the registers of use.
9. The functioning of neural machine translation (NMT) was then discussed: learning from large multilingual corpora, the use of self-attention mechanisms, but also structural limitations such as the absence of true semantic understanding and difficulties with metaphors, cultural references and sensitive contexts. In this context, the figure of the translator emerges as an “algorithmic curator”, who uses AI for speed and efficiency, but critically monitors errors, pragmatic adequacy and ethical responsibility. The challenge was to have the AI generate a three-day tourist itinerary in Spain and evaluate its accuracy, feasibility and cultural sensitivity. The groups thus developed two itineraries, in Barcelona and Seville, according to the assignment.

A multidisciplinary workshop where AI is learned by practising it

The challenge, organised with the support of Dario Nucifora and Veronica Chisari, concluded with the groups' final pitches and a collective discussion on the value of interdisciplinarity. The initiative showed how AI, when incorporated into educational programmes, requires not only technical skills, but also interpretative abilities, cultural sensitivity and a solid humanistic foundation. The meeting in Palermo confirms once again that innovation, especially when it comes to culture, arises where different disciplines learn to dialogue with each other.

The instrumental role of AI and human decision-making

The students recognised how artificial intelligence “can open many doors”, highlighting its potential as a resource. However, during the group work, they experienced how achieving the desired results is not automatic. A crucial element that emerged was the absolute necessity of human intervention (human intervention was really necessary), even after using prompts or initial stimuli for AI. In particular, one of the challenges involved recreating paintings with artificial intelligence, starting from photos of original works. Students had to use different artificial intelligences to compare results and determine which was the most intuitive for the artistic aspect. This activity highlighted the value of different human skills, especially in the context of the humanities. For example, those from artistic or humanistic fields were able to more easily understand what specific instructions to give to AI (such as “brushstroke”, “state of preservation”, or “wear”) to achieve an accurate result, compared to those with a STEM background, who felt more like “technical support”.

The lens metaphor and the necessary education

The most effective summary of this collaborative relationship was provided by one participant who defined artificial intelligence as a “lens”. In this analogy, AI is a tool through which one observes or processes, but “we humans then have the decision-making power because we were ultimately the ones who decided”. This decision-making power is exercised by bringing together different disciplines and specific skills, combining human intelligence to finalise the work. This dynamic led the students to emphasise the fundamental importance of specific training in the use of these tools. The need for “education on the tool, in my opinion, on artificial intelligence” clearly emerged in order to communicate effectively across different disciplines and make the most of AI's potential. This is crucial because AI is understood as a tool that looks to the past, useful for reorganising and synthesising content, but not for producing original content, which remains the strength of human beings.

The key concept is that AI does not replace, but amplifies the human ability to understand and connect, when used with awareness. Like a lens, AI can magnify or focus data, but the direction and interpretation remain in the hands of the user.

Travel diary

We conclude the report on this intense stage with the account of Cecilia Stajano, community coordinator for the Fondazione Mondo Digitale. From the pages of her travel diary:

In generous Palermo, everything is very consistent. An event that is easy to organise, despite its complexity, because the people of Palermo put their hearts and souls into it, with no formalities, little bureaucracy or mannerisms, just substance.

Those who know how to do so make their skills available, and there is a strong sense of pride in being part of this community, in creating culture as a university. There is respect among colleagues, there is commitment on the part of the vice-chancellor, who puts herself on the line first, chooses the challenges and delegates to the right people.

I find this approach and the keen sense of belonging that is also conveyed to university colleagues who come from other regions for a period of their career very fascinating. Even the students feel that here in Palermo they can train and stay. So I would say that this city sets an example, as I like to say.

Palermo has a particularly positive effect on me. I am happy to be in Sicily as a whole, but Palermo always gives me a recurring feeling of connection, a strong human connection. Palermo seems to speak ever louder and prouder, despite the many contradictions and the many things that are still left there... which perhaps one day will be sorted out, like those balconies without floors, yet with their strength, still attached to an extraordinary building, they leave their mark.

Palermo shakes you up, slaps you with different emotions.

Here, there is still the authentic pleasure of meeting on the street, at the bar, in the square, to build a future or to heal the past. Here, everyone is out and about, always. To meet, you go out and see each other, above all. Palermo must have suffered I don't know how much during the pandemic, going against its nature as a natural connector. Because of the work I do and the person I am, being able to so easily and quickly make room for relationships and build something on the spot is a treasure. For me, this is a nourishing, fertile, generative land. It always has been, even when, for extreme reasons, it had to put this gift on hold. I hope that this land full of colours, flavours, smells and souls will increasingly be able to demonstrate its value. We are betting on it, and on this trip we have further laid the foundations for learning from those who, with discretion and strength, protect, honour and preserve it every day.