Title

Sara Dolci - Università di Bologna

The brief that have been assigned to you by Elbert Girón

Crear una colección de utensilios de cocina moderna para chefs Indígenas Oaxaqueños en restaurantes de 5 estrellas, que incluya objetos como tortilladora, metate y cucharón. Esta colección será utilizada para la preparación de comidas como mole, tetela, tamales o yique. Los colores y acabados deben ser similares a los de las artesanías tradicionales de Oaxaca, mientras que los materiales deben reflejar el diseño contemporáneo Mexicano.

Persona profile V1

Feedbacks

I was pleasantly surprised by the results you were able to generate for the persona profiles. Most images of people from Mexico are typically portrayed as white Hispanic, as opposed to the darker skin of people from Oaxaca. The persona's facial features do not strongly resemble Indigenous features, but I think that is more of an issue with Midjourney. I think the individuals in the generated images looked different, which is typical of Midjourney. I did notice that the colorful patterns in the huipil and chef hats did not align with Oaxacan culture, but seemed to be a mix of different cultures throughout Latin America. The kitchenware appeared to be mostly aluminum, which aligns with a 5-star restaurant, but traditional Oaxacan food is usually prepared using clay and steel kitchenware.

The persona details were well-done. The environmentalist aspect was interesting, as sustainable farming and gardening are a part of rural Oaxacan culture but are not typically framed as environmental endeavors. The use of the term Zapotec is accurate for Oaxaca, with Zapotec and Mixtec groups being common in the region. The passing down of recipes through the mother is typical, and grinding corn is a common practice for higher-end restaurants and rural home meals. Walking through local markets is also typical for most people in larger towns and cities in Oaxaca. The holidays mentioned are usually larger celebrations, and while a bit random, are accurate. The desires and worries sections are very accurate for a younger Oaxacan chef navigating high-end restaurants, as indigenous Oaxacan food has only recently been recognized as legitimate in fine dining.

Persona profile V2

Feedback V2

You did a great job with the persona again. I think there are aspects of Machismo (a more conservative view of gender roles) that women, especially in Indigenous Mexican groups, have to deal with that was not represented here. I'm not sure if that is because ChatGPT is not aware of those cultural viewpoints, or if it is offering a more inclusive view in making the Persona. The López Hernández surname is accurate, taking both the family names. The desires and concerns are expected, with the desire to promote undervalued traditional foods and preserving cultural identity being very common with people around this age. The Mexican government has been making efforts around preserving cultural identity, especially with diasporic change due to immigration.

I think you did a great job with Midjourney, as the people look really similar from image to image. What I do notice is that Midjourney continues to struggle with creating a kitchen environment that makes high end and indigenous foods, as seen in the walls and kitchenware. Again, Midjourney struggles with replicating traditional patterns in the clothing but that is expected as Midjourney struggles with accuracy. Midjourney also struggled with creating foods, making something that looks like white rice and unclear vegetables. 

Feedback V2

I think this was a really good attempt at bridging indigenous foods and high-end restaurants. I think your prompts were written well, in a way that allows different elements to come together in a balanced way. Midjourney is still struggling with making accurate foods, something that we both have noticed. I think the image of mole and tamales was great. Some textures, like the texture of the corn husk, were out of place, but I think it is generally still recognizable. They also captured a lot of elements in one image, which I remember was a bit more difficult in the first attempt.

Feedback V2

I like that you expanded on the previous work and included a characteristics, instructions, and care section. That helps add to the product narrative, and was generally accurate. The Metate y Mano was great. I can tell you took the feedback from the last time and worked on it. The combination of traditional forms and contemporary materials was great and tied back to the brief well.

Some discrepancies I noticed were that the Tortilladora looks similar to existing products, but is it not clear how the mechanism would work to flatten the dough. Quesadillas are not able to made with it, but the sopes would. I also noticed that the cleaning section mentions waiting for the Tortilladora to cool down, however it is a press and does not cook or heat.

Again, I think the espumadera is a really successful product. Including details like combiing different technologies, like painting and laser engraving. It really captures the brief of combining contemporary and indigenous together. 

Feedbacks (first concept)

The concepts were great! I felt that they were accurate to the brief.

General feedback is mostly about the balance between materials. In the Metate y Mano, traditionally materials like stone are preferred. However, materials such as aluminum and steel are more food safe and used more often in a restaurant kitchen. Perhaps focusing on material and finish as a way of emulating the traditional materials is an option.

In the Tortillera, the use of color and pattern was very successful. Most tortilla presses in restaurants are very industrial, being made out of iron with little to no decoration. The use of mesquite was a bit out of place. Mesquite wood is used often for Mexican utensils, but a wood like Zapote is more specific to Oaxaca. The adjustable scew mechanism is a good function and is relevant to making tortillas.

The Espumadera was my favorite object. It is typically a plain metal utensil. Incorporating traditional patterns and color is both functional and culturally relevant. 

As Andrea mentioned, thinking about which materials are food safe is important.

Generally, the objects can maybe use a stronger design language to connect each object together to feel like one collection.

Notes and reflections

Hunting Report

Brief

The hunting part of this workshop turned out to be very interesting, thanks to the project path led by the AI, starting from the brief I assigned to my teammate. I approached the first step of brief writing by thinking of something that Chat GPT and Midjourney would 100% mistake with something else, maybe more popular or related to another culture and tradition (i.e. raviole that could be mistaken for ravioli and mustard jam as well). Plus, I focused on a very specific and underestimated part of my culture as well. So once I decided brief’s object, I started generating with ChatGPT a short sentence entering inputs about specific cultural information I wanted the AI to consider.

Input sentence

“Scrivi un brief che non superi i 300 caratteri, dove chiedi di progettare un servizio di consegna delle raviole bolognesi, sapendo che: le raviole sono il piatto forte di una tradizione emiliana di stampo propiziatorio. Il 19 marzo è, infatti, l’antivigilia dell’equinozio di primavera nonché il giorno di San Giuseppe: l’inverno è in procinto di lasciar spazio alla rinascita primaverile e alla ripresa del lavoro nei campi. I contadini bolognesi in questa occasione organizzavano una festa di balli e canti nella speranza di una continuazione dell’anno economicamente più ricca. In quest’occasione le donne erano solite preparare le tipiche raviole ripiene di ghiotta mostarda bolognese. A Fiesso, una frazione di Castenaso, erano utilizzate per addobbare le siepi del circondario: chiunque passeggiava in quei luoghi poteva così godere della dolcezza di questi inusuali frutti.”

“Write a brief that does not exceed 300 characters, requesting to design a delivery service for Bolognese ravioli. Bolognese ravioli are a staple of the Emilian propitiatory tradition. On March 19th, the day before the spring equinox and San Giuseppe’s day, Bolognese farmers traditionally held a feast with dances and songs to welcome the upcoming spring and a prosperous year. Women would prepare the ravioli filled with delicious Bolognese mustard. In Fiesso, a small town in the municipality of Castenaso, ravioli were used to decorate hedges, giving anyone passing by a taste of these unusual fruits.”

The first result was surprising, turning out as a melting pot as cutting of information pieces and sticking them back together. For instance, the “ancient propitiatory ritual” thing about the San Giuseppe’s day was taken up and integrated into the brief as an attribute the service was supposed to have among its main features.

Output sentence

“Ciao, vorrei progettare un servizio di consegna delle raviole bolognesi, piatto forte della tradizione emiliana, con particolare attenzione al 19 marzo, l'antivigilia dell'equinozio di primavera e giorno di San Giuseppe. Vorrei che le consegne fossero fatte con cura e attenzione, come si trattasse di un antico rito propiziatorio, in modo che i clienti possano gustare le tipiche raviole ripiene di mostarda bolognese ovunque si trovino, sentendosi parte della festa dei contadini bolognesi.”

“Hi, I would like to design a delivery service for Bolognese ravioli, a staple dish of the Emilian tradition, with a particular focus on March 19th, the day before the spring equinox and San Giuseppe’s day. I want the deliveries to be done with care and attention, as if it were an ancient propitiatory rite, so that customers can taste the typical ravioli filled with Bolognese mustard wherever they are, feeling part of the Bolognese farmers' celebration.”

Then I kept asking ChatGPT to refine and summarize the brief sentence, adding some details to it, such as features the delivery service should have in terms of sustainability and social innovation. This last part was supposed to be a further challenge to be addressed as well as a hint on issues affecting the rural areas and local communities.

Brief sentence

“Crea un servizio di consegna sostenibile delle raviole bolognesi, piatto tradizionale preparato in occasione del giorno di San Giuseppe. Le consegne si svolgeranno ispirandosi a un antico rito propiziatorio, per far sì che le coppie di innamorati possano gustare le tipiche raviole ripiene di mostarda ovunque si trovino e sentirsi parte della festa contadina che celebra l'arrivo della primavera e dei buoni raccolti. Il servizio dovrà essere sostenibile: produrre zero CO2 e portare innovazione sociale nelle comunità rurali della Pianura Emiliana.”

"Create a sustainable delivery service for Bolognese ravioli, a traditional dish prepared on San Giuseppe’s day. Deliveries will follow an ancient propitiatory rite, allowing couples to taste the typical ravioli filled with mustard wherever they are, and feel part of the farmer's celebration of spring and good harvests. The service should be sustainable, producing zero CO2 and bringing social innovation to rural communities in the Emilian Plain.”

The information given were never too detailed, in order to leave room for misunderstandings and cultural bias. The target user was also added in order to make the following generations of the persona more complex and potentially full of bias to find. 

Persona

Version 1

The first version of the Persona exceeded my expectations, showing me a completely different profile as compared to the target of the brief. At this point I thought the brief was confusing enough to deceive and mislead the AI. Overall, the Persona was quite accurate depicting a farmer of northern Italy and reporting information about his name, age, profession, interests, habits, needs, fears and desires. Even if non-centered with the brief, that represented an occasion to spot more biases. 

The very first bias that got my attention was about gender and age of the Persona. I felt this like a common stereotype widespread in manual-work fields and among rural population. Even if a lot of farmers are men over 40 that can’t still be the absolute truth. In fact, there are also many women that own land and do farming as well as many young people do (nowadays but in the past too). On the other hand, it made sense assuming Marco belonged to a long line of farmers as usually involving family-run activities passed on from generation to generation. 

I felt the name as a forced convention of ChatGPT about Italian names and surnames, assuming “Marco Bianchi” takes up the most common ones. Especially, the surname is well-known in the whole country, even though it doesn’t reflect typically-emilian surnames that come from rural communities. I suggested my teammate some examples of more typical but lesser-known surnames, including Vaccari, Mezzadri, Emiliani, Malagoli. 

Secondly I found some inaccuracies according to the agricultural context, as the Persona worked both as a farmer and a delivery man. That seemed unlikely to match with farmers’ situation in the emilian countryside, who are absorbed almost 24/7 by work in the fields and therefore prefer to sell their products directly at a retail location nearby. 

Furthermore, ChatGPT didn’t make clear what’s the relationship between Marco and the raviole supposed to be, assuming he was also producing them or any of the ingredients. In that case, I noticed further inaccuracies and missing information. I felt unclear which raw materials Marco provided to make raviole (vegetables, fruits, cereals, etc..), so I suggested my teammate to ask ChatGPT for defining better what he grows first, according to the product that had to be delivered. In addition, the raviole-making activity should be an extra business which involved more than one person, so Marco would be able to spend enough time taking care of the crops. 

ChatGPT didn’t specified the marital status, an element that could be important in order to reveal more gender and culture-related biases, as well as other more general information - interests, wishes and fears - about Persona’s life, even if not strictly linked to the brief. 

Looking at the images Midjourney generated, I felt accurate and realistic the face features and the physical characteristics of the Persona, especially for the tanned complexion, the wrinkles and the big hands. The vegetables inside the baskets and the crops in the background were out of place, regarding the raviole bolognesi as the main product to produce and deliver. 

Also the prompt suggested a glitch the AI had while generating, saying mustard greens as typical ingredients of emilian cuisine, as well as breadcrumbs and traditional bread rolls were key details from San Giuseppe’s day celebration. One of the two images showed in the background a church not resembling the gothic-styled churches of Northern Italy.

I was expecting the raviole were mislead by the AI, so I provided the synonym raviole di San Giuseppe my teammate could use during persona refinement and concept generation step, giving him some hints about the dish. 

Version 2

The second attempt led to a total makeover of the persona as the AI generated two profiles in line with brief’s indications. At the beginning, I wrote the brief focussing on a stereotypical target, in order to disseminate the hunting seeds. The young-loving-couple target was a compromise to combine traditional elements of the emilian cuisine and innovative components of today’s needs. In fact, the brief considered new generations as a key-role in preserving cultural heritage. 

Some mistakes, misconceptions and biases have come up, highlighting some gender-role issues. First of all, ChatGPT generated a heterosexual couple leaving out any queer relationship. In the employment section an inconvenient truth emerged about the actual problem in Italy, thus still limiting professional roles according to specific genders. In fact, the teaching role is often related to the female gender (especially when dealing with kindergarten and primary school), while tech jobs are more likely bound to male people. At this point I was curious to know if ChatGPT was able to cross the border generating also non gender-related professionals. 

Again a bias showed up in Persona’s interests, facing some gender issues. This time it was involving something typical about hobbies people usually have in Italy. On one hand boys tend to dedicate themselves to sport disciplines and video games, while girls are interested more in nature, arts and culture on the other. This discrimination is manifesting a consistent gender gap, as well as typical features of the Italian culture. Also there I was wondering how could the AI go further and approach this conventional situation in a different way.

I felt accurate and also far from conservative the habits that were generated, as in several regions of the country there’s gender divide at multiple levels, including in the home environment. For instance, in some cases only girls or women are supposed to take care of the house and cooking as well. Otherwise, Luca and Sofia share the same passion for cooking and trying new recipes, even if both of them are quite busy working during the day.

Wishes and concerns got either positive or negative feedbacks. On one side I didn’t get exactly the cultural heritage the personas wanted to preserve, as there are several subcultures to deal with that haven’t to be seen as one - or at least can’t be considered as a whole across the Emilia-Romagna region. I thought that adding information such persona’s hometown could help identifying a social segment or group and a more specific culture too. On the other side, I found interesting the reason why Sofia and Luca are willing to keep in touch with their roots. I surprisingly empathized with the personas as they feared to lose track of their origins, since I feel connected to my roots through many childhood memories and moments spent with my family doing periodical activities. I brought up some references about particular practices as in-depth insights to generate a few new elements.

The pictures were well generated, showing Luca and Sofia in several situations. They seemed italians as I look to their facial features and also looked so similar in every picture. In my project path, I personally experienced difficulties in generating the same person in more images instead. Some biases came up as I look closer at the details. Firstly both the personas were all dress up as they went to a very special occasion, unlike 30-aged people usually do. Luca was wearing suit and tie while Sofia was supposed to have on high heels and a skirt, according to the prompt. 

I found curious the way ChatGPT describes Basilica di San Petronio in the first prompt, because it doesn’t resemble the real architecture. It kinda looks like a generalization about Gothic style instead. However, the picture didn’t show any church or gothic-styled building, demonstrating how the AI struggles to follow prompt’s contents. 

The second picture surprised me with red walls in the background, reminding me "Rosso Bologna”: a red hue typical of Bologna. Otherwise, I found so confusing the two dishes laying on the table in front of the couple, very different from emilian cuisine. it turned out to be something fried, like tempura. In addition, the raviole ChatGPT was describing in the prompt was completely wrong as I expected. It’s funny observing that ChatGPT gets always confused about raviole as it comes up with new recipes every time, with tomato sauce and cheese seeming good on everything.

At least, I found the third image very detailed and also specific about the place. I appreciate that a small town like Castelveltro di Modena was taken into consideration by the AI. Even if it didn't depict exactly the place, there were some traditional elements in the buildings such as the porches, while the architectural style seems to date back to Renaissance, instead of Gothic period showing different colors and materials.

Concept

Version 1

The first concept generation turned out to be very interesting and overall defined in every detail, following many parts of the brief. I found the idea innovative, especially for the social part, in fact none of the existing delivery services values the traditional emilian kitchen in the region. The bigger mistake the AI made was about traditional dishes description, saying tortellini are related to mustard greens somehow. Also, the concept’s image didn't show any raviole, which was the main product according to the brief. 

I found many interesting elements, such as the local farmers engagement, the support to spread sustainable practices in agriculture, the electric vehicles provided by local businesses and the meal kit made of farmers’ fresh products. In particular, I found in this last one a connection with persona’s need to eat a quick lunch without sacrificing quality.

On the other hand, I felt less the traditional part about preserving cultural heritage. Also there was a lack of information about service touchpoint, such as apps, websites, social networks and promotional materials. The acquiring of electric cars and bikes for the deliveries was supposed to be improved adding information about the system behind the partnerships with local businesses (e.g. defining a sort of win-win strategy that ChatGPT could generate). 

In summary, the feedbacks were positive with just a few inaccuracies to point out. I felt the AI encountered some challenges in respecting every point of the brief, especially in defining dishes characteristics. The service could be effectively doable and work very well in my opinion. The pictures showed various kinds of vegetables that as usual seemed undefined and not so realistic looking twice. 

Version 2

Concept 1

At first sight I liked so much the idea, seeing a highly potential to extend the service on a wider scale, even on the whole country. I noticed that ChatGPT always lose track of the brief as including more products than just delivering the raviole (I don’t see it necessarily as a mistake, but a strength instead). I felt useful the net the service could build to spread local traditions, both inside and outside Emilia-Romagna region. Also, I think that providing a digital touchpoint helps a lot in order to engage younger people - in line with the target profile - and diversify the channels the target is reached through. 

There was a contradiction I noticed between making traditional food accessible to vegan and vegetarian customers and preserving the original recipes as can be hard avoiding contamination and distortion. ChatGPT should considerate the heritage-loss risk, ensuring a way in which both tradition and accessibility are taken into account. In order to make service’s name even more rooted in the tradition I was wondering also the possibility of asking ChatGPT to generate bolognese-dialect words. 

Some information was missing about the service-functioning system. Why would users use that service? Why would grandmothers work in this net? How could customers and producers get in touch with each other? ChatGPT should have considered better means, actors and resources of the service on that, as well as generated some references about benefits for both sides. 

A particular inaccuracy I found out was about the part concerning live video chat, where grandmas could provide a guided tutorial while cooking the meals following the traditional recipes. Even if that was the service’s feature I liked the most, I can see the struggle of elderly population dealing with technology, especially in Italy. For this reason the only use of the online platform as the main touchpoint between customer and producer could raise new barriers causing inefficiencies. A support for grandmas is needed aiming to reduce digital divide and generational gap as well, enhancing accessibility and promoting the relation between innovation and tradition. 

In the image Midjourney generated I could see a lot of ragù that seems to be quite accurate. Some boxes contain unknown food to me reminding barley salad. Also, I noticed some green leaves used as topping ingredients which is unusual for typically-emilian dishes. At the centre of the image there was something that resembles almost completely a lasagna, even if a little bit different.

Concept 2

The second design concept stands out from the others thanks to the union of both awareness and practical approach to get people closer to the tradition, rather than just “consume it”. Here have emerged many features in common with other concepts, like the integration of digital aspects touchpoint to complete the service, as well as the virtual cooking classes idea to engage users as an additional function and to keep them in touch. 

Also here I’d like visualizing better the mechanism behind the service (i.e. how partnerships work, advantages for local farmers to become part of the system, how users can check and pick the kits), including also the way orders are delivered, according to the brief.

I was thinking even if would be better making distinctions between the various dishes/traditions over Emilia-Romagna territories, since there are many cuisines, each strictly attached to its place of origin. As it would be complicated put all of them together without categorization, the AI answered with glitches, without taking into consideration the high diversity of cultural heritage across one region. This could be reductive, leading to further misunderstandings and damages to culture and tradition. 

The picture turned out very well. As usual Midjourney struggles generating people, details or following the prompt at every point. I felt correct including fruit trees in the prompt because fruit-growing areas are widespread in Emilia-Romagna. I’d like to see what Midjourney was capable to generate if asking for the meal kit and the recipe card to be shown. 

Concept 3

In the last concept I would say that the idea brought together communities’ and customers’ needs. I appreciate ChatGPT considered additional food products and drinks that are very typical about Emilia, such as balsamic vinegar and Parmigiano-Reggiano cheese. Also here the digital integration of events and workshops is very effective to spread traditional practices and to keep them alive through the young generations. I felt very consistent the basket weaving mentioned as an example for lab activities as that resulted typical about farmers’ handmade-traditional practices from the past.

As for the previous concepts, I’d like visualizing better the mechanism behind the service (i.e. how partnerships work, advantages for local farmers to become part of the system, how users can check and pick the kits), including also the way orders are delivered, according to the brief.

The picture seemed not enough evoking the concept idea. I felt misleading the picture as it wasn’t depicting the final product of the service but ingredients instead. In order to fix that, I suggested to generate lunch meals and maybe one more image showing an activity/workshop among those mentioned in the concept.

Generally the raviole came out several times in concept descriptions, but were never depicted in the picture or even generated by Midjourney. 

Conclusion

Overall, I would say that the hunting activity turned out to be a key-step to make visible misconceptions, biases and contradictions about emilian culture. Even if raviole didn’t show up a lot in images, I noticed many wrong information were provided by the AI, for example making up brand new recipes including random tipically-italian ingredients. Both ChatGPT and Midjourney won’t be able to generate well-defined elements, such as existing buildings, cities, food and people, basing on general information and clichés especially. Furthermore, gender-based issues came out very often, concerning rural communities situation both locally and nationwide. The profiling activity helped figuring out the main stereotypes, highlighting the links between profession roles or interests/activities and gender. In general, the AI technology is becoming increasingly part of our lives as an everyday tool used for work and leisure. On the other side, we should consider its potentials, boundaries and threats in order to manage this relationship of dependency.