Premise
Using artificial intelligence technology for designing tattoos to help those who are experiencing decision paralysis when choosing tattoos.
Synopsis
Today, too many different tattoo designs can be found all over the internet and on social media; however, like any form of consumption, people who want to get tattoos can be overwhelmed by the choices available and face decision paralysis, which may take them the time and energy to overthink and cause anxiety, so it is essential to think about a solution to help them choose a design that matches their preferences. One potential solution could be creative design using artificial intelligence technology. AI has been successful in many creative areas, such as producing imaginary content and images. In addition, recommender systems are effective for finding the right item of content when individuals are overwhelmed by decision paralysis. For this project, I developed a prototype of an app that generates tattoo designs based on the user’s imagination or preferences and tested its effectiveness with young women.
Substantiation
A tattoo is a form of body modification where ink is injected into the skin to create a permanent design for various reasons, including self-expression, individuality, and association. (Sun, Z.H. et al., 2016). Tattoos are no longer considered a stigma in society, and throughout the past few decades, there has been a noticeable increase in the number of people acquiring tattoos (McCandlish et al., 2023). For instance, in Amsterdam, the number of tattoo shops has increased from 60 in 2009 to 109 in 2019 (Ozcan, 2023). However, the choice of design among too many designs, which are constantly updated and promoted on social media, can be overwhelming for people who want to get a tattoo. According to the choice overload theory, having more options to choose from may lead individuals to encounter difficulty deciding (Scheibehenne et al., 2010). Inability to decide as a result of excessive overthinking is known as decision paralysis. Some studies show that analysis paralysis in various businesses is a result of choice overload (Adriatico et al., 2022). The phenomenon of decision paralysis consists of regret anticipation, overthinking, inaction, and delay and may affect people’s decision process (Manolică et al., 2021). To design the prototype, I chose AI because AI creativity has been making a significant impact in various fields, bringing new ideas and possibilities to individuals (Wu et al., 2021). AI tattoo design tools can help to democratize the design process and enable more people to create custom tattoos that reflect their personal style and preferences. And a recommendation system (RS) would be useful to recommend an item to a user based on given information (Fayyaz et al. 2020). So, my approach involves using machine learning algorithms to identify and recommend a small set of designs that are most likely to meet a user’s preferences. Choosing the best-fitting tattoos for young people is related to some factors, including the design, the tattoo designers’ copyright rules, and the meaning and association with specific groups in society. However, my specific scope in this project is only to research and design a tool to prevent young women’s tattoo decision paralysis so, the other factors were not considered.
Iteration 1
For the first iteration, a python code was written that used an open AI API key to connect to server DALLE 2, which is an AI system that can create realistic images from a description prompt in natural language. And tested with four young women aged 19–24; one of them had tattoos and would like to have more, and the rest of them said they think about getting tattoos. They were asked to feel free about the sentences or phrases they entered into the prompt; however, they had to write the phrase ‘a tattoo design’ at the beginning of the last sentence to make sense in natural language for AI. Then they were asked about their experience.
Insights from the testing first iteration:
• They were overly excited and quickly produced image after image.
• They had to add fine-line phrases in their sentences to get a fine-line tattoo image.
• One of the users had a problem with the interface and asked for a more user-friendly interface.
• They could not produce a tattoo design with politicians and got errors (based on DALL-E 2’s content policy limitations).
• When they reversed the sentences, they got considerably different images. For instance, these two images generated considerably different.
sunset tattoo design
• Tuning prompts are important. According to open AI, better prompts could improve the performance of the images to give good results about 50% to 65% of the time, but often not beyond.
Iteration 2
For the second iteration, A user-friendly interface was created. And a tattoo design with a phrase adds to the prompt as a default text in case users will forget to type it. They could even totally omit this phrase or put it at the end of the sentence.
Insights from the testing second iteration:
• Although they were free to choose one specific design, they created one based on their preferences, and it was still difficult for them to choose.
• After testing five or six images then they felt a lack of imagination and took time to think of new sentences.
Iteration 3
Since recommendation systems offer suggestions to navigate users to the items which are most likely of interest to them efficiently (Burke, 2007; Gorgoglione et al., 2019). For the third iteration, I used the non-personalized recommendation system method because the App was tested with random users so, there was not any form of collaboration between users that they could suggest together. with an API, I got a CSV file dataset of all objects of the New York Metropolitan Museum of Art. And then selected only the paintings titles (5926 titles) column from all objects and save it into a CSV file. And then a python code was written to generate random titles.
Insights from the testing second iteration:
• One of the users suggested it would be great if they are recommended by specific categories. For instance, nature, animals, and historical or cultural symbols.
Conclusion
This app was written for young women who have difficulty choosing when they have a variety of tattoo design options and tested with potential users who were faced decision paralyzed when they offered too many products. overlay, users were satisfied with the creative AI art and saved most of the tattoo design for themselves. possibly they will use them as a tattoo design. However, there is a lot of room for improvement in the prototype and the interface for the future. I will work on more machine-learning language to find better solutions to tune prompts.
References
Sun, Z. H., Baumes, J., Tunison, P., Turek, M., & Hoogs, A. (2016, December). Tattoo detection and localization using region-based deep learning. In 2016 23rd International Conference on Pattern Recognition (ICPR) (pp. 3055-3060). IEEE.
McCandlish, C., & Pearson, M. (2023). Tattoos as symbols–an exploration of the relationship between tattoos and mental health. The Journal of Mental Health Training, Education and Practice.
Ozcan, U. (2023). The health of tattoo artists. EUR J ENV PUBLIC HLT. 2023; 7 (2): em0131.
Scheibehenne, B., Greifeneder, R., & Todd, P. M. (2010). Can there ever be too many options? A meta-analytic review of choice overload. Journal of consumer research, 37(3), 409-425.
Adriatico, J. M., Cruz, A., Tiong, R. C., & Racho-Sabugo, C. R. (2022). An analysis on the impact of choice overload on consumer decision paralysis. Journal of Economics, Finance and Accounting Studies, 4(1), 55-75.
Manolică, A., Guță, A. S., Roman, T., & Dragăn, L. M. (2021). Is consumer over choice a reason for decision paralysis? Sustainability, 13(11), 5920.
Wu, Z., Ji, D., Yu, K., Zeng, X., Wu, D., & Shidujaman, M. (2021). AI creativity and the human-AI co-creation model. In Human-Computer Interaction. Theory, Methods, and Tools: Thematic Area, HCI 2021, Held as Part of the 23rd HCI International Conference, HCII 2021, Virtual Event, July 24–29, 2021, Proceedings, Part I 23 (pp. 171-190). Springer International Publishing.
Fayyaz, Z., Ebrahimian, M., Nawara, D., Ibrahim, A., & Kashef, R. (2020). Recommendation systems: Algorithms, challenges, metrics, and business opportunities. applied sciences, 10(21), 7748.
Burke, R. (2007). Hybrid web recommender systems. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 4321 LNCS, 377–408.
Gorgoglione, M., Panniello, U., & Tuzhilin, A. (2019). Recommendation strategies in personalization applications. Information and Management, 56(6), 103143.