As I work my way towards my thesis topic, I am struggling with the feeling that “UX design and AI” is both too broad and too specific. I also wonder if it’s really the topic for me, and I have been asking myself why I chose it in the first place, what exactly interests me, and most importantly, what am I passionate about?
I initially chose to research AI because I felt I had very little knowledge of it, and I wanted to empower myself, and others by extension, to understand and use AI. Last semester, I floated the idea of designing an app using as much AI as possible, and writing a thesis on the process and outcome. But after a summer of reflection, that concept isn’t sitting quite right with me.
My initial interest in AI was driven by my love of people and of helping others, not by my love of technological innovation. Although studying interaction design means that I possess the prerequisite of a fascination with technology, my strongest motivators come from a love of humanity and human connection rather than innovation for innovation’s sake.
If I were to design an app using as much AI as possible, who am I helping? What am I accomplishing? With AI developing so rapidly, surely my work will be outdated before I am even finished. This led me to consider other ways to use my now accumulated AI knowledge. Perhaps a guide for designers, maybe in the form of an app, on how to coexist peacefully with AI? I can’t say I have come even close to a conclusion, but I am keeping my core motivators front of mind as I move forward, with the hope that they will guide me to a topic that both poses a challenge and aligns with my values and interests.
This week, I finally cracked open and read Hooked by Nir Eyal from cover to cover. When looking for inspiration, I tend to gravitate to the “basics”, with the intention of building from there and hopefully discovering something on the way. At this point in time, I would like to build my library of UX Design must-reads, continuing with The Design of Everyday Things by Don Norman and 100 Things Every Designer Needs to Know About People by Susan Weinschenk.
While I took extensive notes as Nir Eyal laid out the four stages of the Hooked mode, what interested me most was in the last 50 pages, in a chapter entitled “What Are You Going to Do with This?”, which was, coincidentally, also my question at the time. In this chapter, Eyal discusses manipulation in technology and innovation, and dives into how to use the Hooked model ethically. The author discusses the relative newness of technology and our unavoidable ignorance of possible future consequences, but also argues that less than 1% of users develop a detrimental addiction to anything, and that, excluding vulnerable groups, users are responsible for themselves and are able to self-regulate their behaviours.
Using his manipulation matrix, Eyal sorts creators into four groups: Dealers, Peddlers, Entertainers and Facilitators. Eyal argues that “manipulation doesn’t always have a negative connotation”. The book later cites The Bible App and Fitbod as examples of companies that used the Hooked model to build healthy habits.
Nir Eyal’s discussion on ethics in design relates strongly to conversations on ethics in AI. In both sectors, there are advocates for an ethical code of conduct. In both sectors, manipulation is a front page topic of discussion. In both sectors, there is opportunity for humans to work with machines either for or against other humans. In both sectors, we must work with clear intentions to use the tools we have FOR and WITH people.
Hooked was a very different read from what I expected. Like the author expects, I was uncomfortable with the idea of “a cookbook for mind control”, but instead found thoughtful reflections on meaningful design that I will bring forward throughout the process of writing my thesis, and beyond.
In preparation for writing the master’s thesis next semester, we were asked to review a completed thesis from a past year. I took a look at Money Talk by Ines Mahmoud, and found a lot of inspiration for my own work in the content, structure, and heart of the thesis.
Money Talk analyzes the issue of the financial knowledge disparity between men and women and proposes an intervention in the form of an app. The thesis is broken down into 5 parts, plus the conclusion and reference list. The introduction includes the relevance of research, problem statement, state of research, goal & hypothesis, and approach. The main body of research is titled “Money, Money, Money”, and highlights various belief systems around money, money as a taboo topic, psychological phenomena such as human-money interaction, mental accounting, and materiality of money, and finally the relationship between women, finances, and money, including topics such as pensions, part-time and care work.
The third section is dedicated to expert interviews. The author first interviews Nicole Katsioulis, a political foundation employee and founder of blog “Geldmarie”, which empowers women to understand finances and become financially independent, sustainable investors. Nina von Gayl, the third interviewee, is the curator for Erste Group’s Financial Life Park, and the third interviewee, Maria Mann, is the CEO of a Munich-based start-up called Financery, a “pragmatic tool for women to invest and grow their wealth”.
Following the interviews, the author titles the fourth section “Auto-Ethnographic Observations”, where she dives into her own thoughts, feelings, perceptions, and histories around money and financial management. I found this section particularly interesting because of the method’s inherent bias, which the author addresses on the first page with the following quotes: “Authoethnography is a set of research that seeks to describe and systematically analyse (grafie) personal experience (auto) in order to understand cultural experience (ethno)” (Ellis 2004; Holman Jones 2005). “It challenges canonical practices of conducting and presenting research (Spry 2001) and treats research as a political and social act” (Adams & Holman Jones 2008). I am interested in looking further into this unfamiliar and seemingly unorthodox research method, as I greatly admire the “human-ness” it brings to the work.
The fifth section of the thesis is dedicated to the practical work, where Mahmoud details her process designing the app, Money Talk, and presents the outcome in a creative and engaging format, utilizing cutouts to showcase the screen design. In this section, Mahmoud details each step of the design process, and walks the reader through every consideration of the screen designs, resulting in a complete and comprehensive analysis.
I really enjoyed sifting through Money Talk. I found myself greatly inspired by the content, structure, and heart of this thesis, and already plan to reference it later on as I continue developing my own thesis. I found Mahmoud’s approach extremely well-organized and thoughtful. Stylistically, the work was beautifully presented, readable but also fun and engaging. The inclusion of research, expert interviews, and practical components was a winning combination in my view, and a format I would consider emulating, or being inspired by, for my own work.
On October 11th, 2023, I attended a talk entitled Mastering the Emotional User Experience by Bill Albert, SVP of Global Customer Experience at Mach49 in Silicon Valley. Albert began the talk by noting that the emotional user experience was the topic of a recently added chapter in the newest edition of his book, Measuring the User Experience: Collecting, Analyzing, and Presenting Usability Metrics and something his company is now just beginning to touch on.
Why should we care about measuring emotional UX? Albert cites improving products/services beyond usability, achieving KPIs (loyalty, satisfaction, etc.), aligning with brand strategy, and gaining competitive advantage as positive outcomes for considering emotions in UX research.
Albert utilizes scales of “Arousal” and “Valence” to classify emotions. Arousal classifies the degree of excitement or engagement from “calm” to “excited’. Valence measures the “goodness” or “badness” of a response from negative to positive. Albert argues that only a few emotions are truly relevant to UX design. He includes engagement, confidence, trust, frustration, affect, and stress, but also notes that these are highly contextual. As an example, Albert shared a user testing video for a participant who was laughing at the poor functioning of a website. The face tracking software registered her smiles as a “joyful” response, where in reality she was amused at how bad the website was. Albert also notes that UX testing usually elicits a “weak signal”, or low emotional intensity, and is thus not as obvious to measure compared to high emotional intensity events such as a roller coaster ride, a great first date, or a team winning a big game. The example of the user laughing at the poor functioning of the website, and that response being registered as “joy” is known as “noisy data”, or data that has to be sifted through and removed by a person, not AI. Albert points out that we don’t all have the same definition of each emotion, and we have to also consider cultural and language-based differences – this all results in “noisy data”.
As we begin the semester, I am thinking about how I can narrow my research and get closer to my final thesis topic. Bill Albert’s talk was very relevant to my past research, as he exposed the shortcomings of AI in understanding human behavior, and the prevailing importance of having real people test, analyze, and summarize data. This “impulse” was a great way to get the ball rolling for this semester of Design and Research, and will provide “food for thought” moving forward.
Today I’ll be talking through my prototype for design and research this term. My current research is exploring AI in UX design – asking how can UX designers use AI to streamline and improve their work, as well as finding where AI can’t do the job, due to gaps of knowledge in the source material, or other shortcomings of the software.
For my prototype this semester, I decided to begin the UX design process with an app I had previously created, but this time using as many AI programs as possible. For now, we are specifically looking at user research and persona development.
The app I am working with is called “ProspHER”, which is a job-seeking platform for underprivileged women. The app can be used to find work that fits your experience and skill level, as well as your availability. It specifically highlights unconventional working hours for maximum flexibility when considering caring for children or elderly family members, work that is often undertaken by women.
For my user research, I worked with three scholarly articles on various topics surrounding women’s employment. I ran these articles through an AI PDF reader, and could then ask the AI further questions about the articles.
After conducting this simulated user research, I used an AI tool to generate a persona based off a short description of my app concept. The original persona was surprisingly helpful, and then I added information from my user research to fill it out more.
Important things to highlight in this process are that these AI tools only work well if you know how to use them – meaning that you need to know what questions to ask and how to ask them. You also need to keep in mind, with all AI tools, that these programs are simply remixing information and images that already exist somewhere. This means that our biases are inherently baked in to the results, and need to be watched for with a very careful eye.
In the next semester, I am excited to get deeper into existing literature on AI. I want to see what has already been done to implement AI into the UX profession on an official basis, and to continue experimenting with how much AI I can involve in the UX process.
I began the prototyping process by conducting user research and building personas with the help of AI. I gave ChatPDF several articles on the topic of the experience of women in the workforce, and asked questions in a pseudo-interview style. Additionally, I described the concept of my app, ProspHER, to an AI that generates user personas, and edited the result to include information gathered from my “interviews” with ChatPDF.
Using these AI tools made the user research and persona building process fast and smooth. With ChatPDF, I could “read” many more articles, processing a lot more data in a shorter period of time than I would normally be able to. The user persona AI instantly generated a very usable persona template. These tools sped up my work process enormously. There is some discomfort in putting complete trust in ChatPDF’s interpretation of the articles, however after questioning the AI, I found its answers consistent and unbiased. It often repeated the same information no matter how I phrased a question in an attempt to “lead” the answers, convincing me that the AI doesn’t expand beyond the source material.
Below is a user persona that I developed with these AI tools. The persona is created for my app, ProspHER, a tool for underprivileged women looking to find jobs that suit their availability and experiences. I have also included screenshots and transcripts of the “interviews” with ChatPDF and various articles.
The original persona generated by the User Persona AI tool
Persona with adjustments from ChatPDF “interviews”
Transcript from ChatPDF_Gendered Laws and Women in the Workforce by Marie Hyland, Simeon Djankov, and Pinelopi Koujianou Goldberg
PDF: Dear Reader,
Welcome to this insightful PDF file on gendered laws and women in the workforce. In this article, Marie Hyland, Simeon Djankov, and Pinelopi Koujianou Goldberg present a comprehensive analysis of gender discrimination and its impact on women’s economic opportunities. By utilizing the Women, Business and the Law database, the authors shed light on the persistent gender inequalities that exist, particularly in terms of pay and treatment of parenthood. They also highlight the positive correlation between more equal laws and favorable labor market outcomes for women.
Example questions you may have about this PDF file:
How does the World Bank’s Women, Business and the Law database contribute to understanding gender discrimination in the workforce?
What are some of the key findings regarding gender inequalities in pay and treatment of parenthood?
Can you provide more details on the correlations between equal laws and labor market outcomes for women?
We hope you find this PDF file informative and thought-provoking. Enjoy your reading!
Best regards, [Your Name]
Me: How does the World Bank’s Women, Business and the Law database contribute to understanding gender discrimination in the workforce?
PDF: The Women, Business and the Law (WBL) database measures equality of economic opportunity under the law between men and women in 190 economies, for 50 years, from 1970 until today. It focuses on legislation that may impact a woman’s access to employment and entrepreneurial activity, attempting to capture inequality in legislation throughout the duration of a woman’s working life, from the time she can enter the labor force through to retirement. The information on the legal environment in each economy is collected through the collaboration of legal experts based in the World Bank with local experts, including lawyers, judges, civil society representatives, and public officials. The database aims to fill the knowledge gap on the relative severity and evolution of legal gender discrimination around the world. [T1, T3]
Me: What are some of the key findings regarding gender inequalities in pay and treatment of parenthood?
PDF: According to the Women, Business and the Law database, the largest gender inequalities are observed in the areas of Pay and Parenthood. Women are most disadvantaged by the law when it comes to the compensation they receive and to their treatment once they have had children. The authors of the PDF file document large and persistent gender inequalities, especially with regard to pay and treatment of parenthood.
Me: Can you provide more details on the correlations between equal laws and labor market outcomes for women?
PDF: The PDF file highlights that there are positive correlations between more equal laws pertaining to women in the workforce and more equal labor market outcomes for women. Specifically, countries with more equal laws tend to have higher female labor force participation rates and smaller wage gaps between men and women. The authors also mention that results from OLS regressions show that a higher score in the Women, Business and the Law (WBL) index, which measures legal gender equality, is correlated with better labor market outcomes for women, such as higher participation in the nonagricultural labor market and a smaller gender wage gap. However, it is important to note that these correlations do not establish causality.
Me: what are some of the challenges women face when trying to enter the workforce
PDF: The PDF file does not explicitly mention the challenges women face when trying to enter the workforce. However, it does discuss the existence of gender inequalities in the legal environment that may impact a woman’s access to employment and entrepreneurial activity. These inequalities can create barriers and obstacles for women when they are trying to enter the workforce. Additionally, the file mentions that women are particularly disadvantaged in the areas of pay and treatment of parenthood, which could also pose challenges for women seeking to enter or remain in the workforce. [T1, T3]
Me: What are some of the gender inequalities in the legal environment that may impact a woman’s access to employment and entrepreneurial activity?
PDF: According to the PDF file, some of the gender inequalities in the legal environment that may impact a woman’s access to employment and entrepreneurial activity include:
Pay Inequality: Women often face discrimination in terms of the compensation they receive for their work. The file states that women are most disadvantaged by the law when it comes to the compensation they receive.
Treatment of Parenthood: The legal environment can also impact how women are treated once they have had children. The file mentions that women are most disadvantaged by the law in terms of their treatment after becoming parents.
Access to Credit: Laws governing women’s ability to access credit can also be a factor. The Entrepreneurship indicator in the file mentions legislation mandating nondiscrimination in accessing credit as an important aspect of women’s ability to start and run a business.
Contract Signing and Business Registration: Restrictions on a woman’s ability to sign a contract, register a business, or open a bank account can also hinder their entrepreneurial activity. The file provides examples of situations where a married woman may be required to obtain her husband’s permission to engage in these activities.
It is important to note that these are just a few examples mentioned in the PDF file, and there may be other gender inequalities in the legal environment that impact women’s access to employment and entrepreneurial activity. [T1, T2]
Me: is the data in this PDF drawn from a diverse set of women, in different countries and economic circumstances?
PDF: Yes, the data in the PDF file is drawn from a diverse set of women in different countries and economic circumstances. The file mentions that the Women, Business and the Law (WBL) database, which is the primary source of data for the study, covers 190 economies and over 50 years of data. The WBL database collects information on laws and regulations that affect women’s economic opportunities in seven areas: accessing institutions, using property, getting a job, providing incentives to work, building credit, going to court, and protecting women from violence. The file also notes that the study includes economies at all income levels and from all regions of the world. Therefore, the data used in the study is drawn from a diverse set of women in different countries and economic circumstances.
Me: does the PDF discuss difficulties faced by women with less education when trying to enter the labour market?
PDF: No, the PDF file does not specifically discuss the difficulties faced by women with less education when trying to enter the labor market. The focus of the file is on gender inequalities in the legal environment that may impact a woman’s access to employment and entrepreneurial activity. While it acknowledges that there are factors beyond legislation that affect women’s behavior and outcomes indirectly, it does not specifically address the challenges faced by women with less education in entering the labor market.
Me: does the article discuss the impact of higher education on a woman’s experience in the job market?
PDF: No, the PDF file does not specifically discuss the impact of higher education on a woman’s experience in the job market. While it acknowledges that there are factors beyond legislation that affect women’s behavior and outcomes indirectly, it does not specifically address the impact of higher education on women’s job market experiences. The focus of the file is primarily on gender inequalities in the legal environment and the relationship between legal reforms and women’s economic opportunities.
As an exercise, I uploaded two scholarly articles on the topic of women in the workforce to ChatPDF. The AI tool read the documents and was able to answer my questions. I was shocked by the usefulness of this tool, and will be using it again with the development of my prototype for this semester.
Below are screenshots from the ChatPDF conversations.
A combination of ChatGPT & bing that answers any question with sources and links.
DocuAsk & PDFgpt
These tools read any uploaded documents and generate summaries and notes, and are able to answer any questions regarding the document in a chat function. These tools would be useful for user research, analyzing data and quickly pulling the important information.
Uizard & Qoqo
“This tool [Uizard] uses advanced algorithms to generate unique design ideas based on the user’s input, making it an excellent choice for designers who want to save time and stay inspired ” – Perplexity.ai
Qoqo generates personas, user journey maps, and interview questions. [Behind a paywall]
Attention Insight
Offers predictive eye tracking in the form of heat maps on your designs.
For the remainder of this term of Design & Research, I will develop a mobile application using as many AI tools as possible, and document my experiences here. In order to test the limits of these tools, I will be working with subject matter than runs at the edge of most data sets – the experiences of underprivileged women.
The app is called ProspHER, and was ideated and designed in partnership with Nagyija Bogás, for our course App Design this semester. I will expand on initial designs created in Figma with copywriting and image generation, as well as carry out user research and testing – all with the help of various AI tools.
Below are several screens from ProspHER. The app works as a job search platform for women with no or little formal training, and/or unusual working hours. Users can search for jobs based on their available hours, interests, and skills. They can quickly build a CV and add past job experiences easily from the ‘history’ tab.
After spending one semester diving into all AI has to offer us as designers, I am undertaking a slight narrowing in scope. In the second semester of design and research, I want to focus my research efforts towards my professional areas of interest. As I would like to work as a UX designer following the completion of the degree program, I am very interested in how AI will, can, and is affecting UX design.
In the initial exploration of this topic, I found two talks given by usability consultant and human computer interaction researcher Jakob Nielsen. In the first, Nielsen was asked about the relationship between AI and UX, and how UX designers can get more involved with AI. Nielsen remarked first that many things are being done with AI now “just because they can [be]”, not necessarily because they are needed. According to Nielsen, this is reminiscent of chasing after the train, or trend, but the use of AI in design is only “good” if it solves a human need in a better and/or faster way. Additionally, exposure to substandard AI products can leave people with a bad impression of AI, who will then be more reluctant to use similar products in the future. For this reason, Nielsen says, it’s better to wait until an AI-integrated design feature is done well before releasing it, rather than hopping on the bandwagon. We must always ask, “What does this actually do for people?”, a sentiment Nielsen believes is lacking in many UX-AI projects.
One aspect of AI in UX that Nielsen is excited about is the possibility of AI becoming proficient at knowing what people want, vs. literally interpreting what they say. Many of us have now had the experience of asking ChatGPT to help us analyze or generate code, or have asked DallE for imagery that never turns out quite as we imagined. Currently, successful use of AI in design requires one to know exactly how to ask for what you want, but Nielsen envisions a future where AI can interpret our wants better than we can communicate them, and “do what I want, not what I say”.
What was particularly interesting to me in these two talks was Nielsen’s predictions for the future of UX work in an increasingly AI-dominated space. In great contrast to many doom and gloom “the robots will take our jobs” positions, Nielsen believes that AI will make UX designers more productive and improve the quality of our output, leading to more jobs and more good UX design being done. More output generally equals more money to be made, and the growth of the UX design field itself. Some examples of AI use in UX include using AI to transcribe user interviews, or to conduct them and analyze the results for points of interest. AI can also comb through massive amounts of data for statistics and points of interest. On the visual side, AI can produce a design draft that is then edited by the designer (we already saw this last term, with Adobe Spark). Nielsen notes that if something can be done more cheaply and easily with AI, then that’s what people will do.
