Project Overview
In the spring of 2024, I enrolled in a course at Carnegie Mellon University School of Computer Science: Persuasive Design. The object of the course was to learn to integrate persuasive psychological techniques into the design of a semester-end project deliverable.
In an effort to empower ordinary people and pursue a solution oriented towards the well-being of individual users, we considered several possibilities. After narrowing our concepts down through secondary research, we determined that there may be a Product Opportunity Gap for an application oriented towards making users more financially responsible.
As the result of secondary research and expert consultations, we found strong evidence to suggest that consumer spending and emotional state are correlated. Given the complexities of this relationship, past academic scholarship and various prototype testing, we determined that there was a potential societal need to develop a product that solved a reframed hypothesized problem statement:
Emotions associated with negative valence, particularly loneliness, are associated with compulsion to spend on products that we associate with positive valence or pleasure-seeking.
However, building new products can be a winding road, and it took us a long time to reach this problem statement, having originally just wanted to improve financial habits.
The Problems
As a result of the often non-linear path in design, we started with our initial problem statement being:
How might we create something that develops strong financial hygiene by employing persuasive design techniques?
This was a challenging problem. Given the evidence that American consumers overspend, how could we convince them to change not only their financial decisions, but their lifestyles? Given the structure of our society and economy, consumerism is necessary for survival. How can we moderate this to a healthy level that not only improves their current financial well-being, but also empowers them to reach their goals and attain financial self-actualization?
The Proposed Solution
We considered a series of persuasive techniques to address this problem. However, given our problem statement and the new findings that we pulled from academic research, expert consultations and prototype testing we determined to use the following:
Obfuscation
Social Proof
Self-Affirmation
Self-Determination
Priming
Based on the critique we received and results from our prototype testing, we pivoted from our initial plans to build a savings tool to a much more immersive experience. Specifically, we pivoted to building an augmented reality game, where a user builds a bond with a virtual dog that joins them in their real-world environment. When users made savings decisions, they received rewards in the game, which they could then trade in to improve their pets.
Without taking a mixed-methods approach to researching our initial problem space of saving money, we would not have been able to design and build such a persuasive, immersive experience.
The Process
Literature Review:
To start, we began by first conducting background research to understand spending habits. We did this by examining various academic survey reports with respect to consumer behavioral patterns and their causes. It was during this initial research that we came across past scholarship that had researched how emotions and consumer behavior were connected. One finding stuck out to us: Emotions of negative valence are associated with consumption of goods that the individual associates with emotions of positive valence. Simply put, when we are feeling bad, we tend to spend on goods or experiences to make us feel better. Another study got even more targeted towards the root cause of this phenomenon, suggesting that loneliness is particularly correlated with overconsumption.
Expert Consultation:
Following this initial literature review, we then progressed to the point where we had narrowed our scope enough to start talking to experts in the field. Leveraging the resources of our university, we spoke to Dr. Jeff Galak, a professor at CMU specializing in consumer behavior and decision making.
Critique Causing a Pivot:
Following a critique session with our classmates, teaching assistants and professors, we then decided to pivot to design a game that obfuscates the original persuasive intent of our project (saving money) through a mental well-being application focused on companionship. This approach allowed us to obfuscate our original goal, while also ostensibly treating one of it's key causes. During this critique, we were recommended using Social Proof as a potential persuasive technique that might help users improve their financial hygiene by competing with other players.
Redefining Our Solution:
In order to use obfuscation this way and build emotional connection, we decided to build an augmented reality mobile game where users took care of a virtual dog. Whenever users saved money in a given period of time, they would be rewarded by in-game currency or points that could then be spent to take care of their augmented reality pet. Leveraging this high-level game design strategy, we felt that we could effectively use Obfuscation as a persuasive method.
Users could then customize their pet, enabling Self-Affirmation as another persuasive technique that we could employ. By awarding points and creating badges, we could also foster competition amongst friend groups, having them compete with one another to score more points or personalize their dog more. This allowed us to also use Self-Determination and Social Proof as persuasive techniques in our design.
Research Through Design:
We then synthesized our findings across our literature reviews and interview. Following a critique session with our peers and professors, we then jumped into prototyping in order to run user studies and rapidly iterate towards a solution.
Bringing AI Into the Research Room:
In order to rapidly iterate, we brought AI assistants into the research room, leveraging image generation models and LLMs to enhance our prototypes and improve our efficiency in developing new prototypes. This empowered us, allowing for rapid iterative prototyping to incrementally improve and test our solution.
Structured Interviews:
Once our prototypes were developed in each iteration, we recruited 15 interview subjects over three rounds of advanced prototyping. To get the feedback that we needed to improve the product, persuasive techniques and validate our concepts, we used moderated user interviews leveraging the prototypes as artifacts to solicit more meaningful feedback.
Prototypes
Low-Fi Prototyping and Concept Validation:
Given that most of our persuasive techniques that we wanted to employ were dependent on building a bond and engaging with an augmented reality pet, the most risky assumption was that people would develop this bond and make these sacrifices at all. As such, our first low-fidelity prototype (visible below) was just to validate the concept that users would desire a tool to help them save and if they would be more likely to save if it was for the well-being of an augmented reality pet. Validating this concept early was critical so that we could start the engineering work to develop the full product the following week.
Low-Fi Prototype Findings:
In general, our low-fidelity prototypes were very well received, with four out of five participants stating that it would encourage them to be more likely to save on discretionary purchases. Additionally, we got very positive feedback on the augmented reality pet, with interview subjects asking us to improve our prototype to "bring the pet into the world with them" more realistically. To us, this affirmed our persuasive approach and AR concept, which allowed us to begin developing the product from an engineering perspective.
Mid-Fi Prototyping, Early Development and Usability Testing:
Following the feedback from users that they would feel more emotionally bonded to their pet, and thereby more likely to sacrifice consumption for it, we improved the fidelity of the pet. Additionally, we modified the AI-generated paper prototypes to emphasize the augmented reality more distinctly. We also began 'Wizard of Oz' tests on LLM and NLP capabilities when it came to providing self-affirmation as a persuasive technique.
Mid-Fi Prototype Findings:
After our mid fidelity prototype testing, we again received very positive feedback with respect to "bringing the pet into the world" from users. We then asked what we could do to enhance their emotional bond to the pet. We got some interesting insights from participants, who generally mentioned that being able to personalize the pet, name, dress or train the dog would make them closer to the animal. This affirmed our choice to build self-determination as a persuasive technique into our prototype.
Additionally, users also liked the added component of conversational interactions that were afforded by using LLMs as a design material to improve the user's experience. Self-affirming messages of support and affirmation on non-finance related topics, such as "Thanks! You're such a great dog dad!" generally drew smiles and a sense of delight from users. This provided further support that self-affirmation was a powerful persuasive technique for us.
Interview subjects responded very well to the augmented reality interactions that we prototyped. They exhibited an overwhelmingly positive reaction to the augmented reality tricks that the dog could perform in the palm of their hand, with their own reality as the back drop. Users mentioned that they would be willing to forgo expenses, possibly even daily, if they knew that the rewards for it would allow them to train their dog on more advanced augmented reality tricks. To us, this solidified our decision to use augmented reality as a modality for design. It also crystallized the strength and potential that self-determination, self-affirmation and obfuscation could have as persuasive forces.
Lastly, one interview subject suggested, of their own accord, to be able to interact with friends' pets who were also on the app and letting their dogs play together with the user. To us, this also confirmed that social proof would be a critical component to modifying user financial behavior and adoption of the application more broadly.
High-Fi Prototyping, Advanced Development and Usability Testing:
After another round of structured interviews, we further refined the paper prototypes and then built out an initial Unity application with an augmented reality pet for play-testing with users.
This basic Unity application required cross-functional collaboration between our team of one designer, two software engineers and a graphic design artist for 3-D modeling. Together, it required the use of a variety of design and development tools, including Figma, Unity and Blender to build the final prototype. Despite these obstacles, we overcame them as a group and built out a full Unity application for testing and demonstrating to the public. Now, the prototype could live within a standalone mobile application, allowing for more advanced testing.
High-Fi Prototype Findings:
After another round of structured interviews, we further refined the paper prototypes and then built out an initial Unity application with an augmented reality pet for play-testing with users.
This basic Unity application required cross-functional collaboration between our team of one designer, two software engineers and a graphic design artist for 3-D modeling. Together, it required the use of a variety of design and development tools, including Figma, Unity and Blender to build the final prototype. Despite these obstacles, we overcame them as a group and built out a full Unity application for testing and demonstrating to the public. Now, the prototype could live within a standalone mobile application, allowing for more advanced testing.
The Results
A Usable Augmented Reality Game Built on Unity:
Given that this was my first time building a game and working with augmented reality, I'm very proud of the final deliverable and the fact that it was a legitimate, independently built 0-to-1 augmented reality product from scratch. Prior to beginning the course, this was a feat that I would not have thought achievable given my experiences and skills coming into the project.
What I Learned:
This basic Unity application required cross-functional collaboration between our team of one designer, two software engineers and a graphic design artist specialized in 3-D modeling. I filled the role of Product Manager, steering these moving parts towards a common goal. I built upon my existing experience in a product role to manage this cross-functional team, but encountered new challenges along the way in building my first augmented reality game.
These challenges taught me the importance of managing a cross-functional product development teams with a hybrid approach, leveraging the best of both agile and waterfall methodologies. Additionally, I learned a lot about the limitations and capabilities of imbedding persuasive psychological methods within the design of a game.