Can AI/ML Augment the Design Process?
Hi, my name is Tobi, and I’m an avid sunset chaser. I’m also a master’s candidate for Experience Design. This article is my way of positing questions / inviting thought and(or) skepticism, and new ideas into my design process for my thesis.
The Design Process
There isn’t a homogenous process followed by designers. Everyone and every company has subscribed to a process that fits in within their design values. But for the purpose of this article, lets assume the standard design process goes as follows:
‘Empathize → Define’ is where we get all the insights about our potential user. We want to know what drives their behaviors, things that matter to them, and how best to create something that enhances their lives. Some answers to these questions are grounded in ethnographic, qualitative and quantitative research. This part of the process produces user personas and journey maps as artifacts.
Now let’s welcome artificial intelligence (AI) and machine learning (ML) to the conversation.
Josh Clark poses this question: “how can designers know where to deploy artificial intelligence?”
I’ve discovered is that AI and ML usually enter the playground in the last three phases, and potentially flow into the final product (digital or physical). AI is being used to augment the experience for users and the process for some designers. A product like Google’s Clip camera helps you focus on the moment so you don’t have to focus finding good photo angles. On the flip side, Airbnb has experimented with generating designs from low-fi wireframes using machine learning’s vision capabilities.
So what seems to be the reason it isn’t present throughout the process? — this is not suggesting we need to start with the ‘tech solution’ then retrace our steps to the ‘problem’.
User Personas and User Stories
A persona is an ideal user of a product or service. Personas have goals and ambitions. As designers, we use personas to immerse ourselves in the minds of users, this in turn helps to learn how different people will experience our products.
Personas contain things like demographics, psychographics, interests/quirks. In Design is Storytelling, Ellen Lupton says it best,“the most valuable personas are based on observing real people.” Personas also appear in user stories, which give more context into the way products will be used in their day to day lives.
As designers we want to eliminate assumptions or perceptions (‘we think our user is 27y/o, runs marathons…’), which is why understanding a user’s needs is the crown jewel of a successful product. This is where AI/ML might have new potential to influence the process — creation of user personas and user stories. I believe if this is incorporated in our workflow, it could lend itself to design use cases/challenges we may not have thought of — more of this to come in a later section of this article.
Framework
This is a proposed framework for how a platform to create personas would work. Combining public record information, market segmentation data, and social media data, we could potentially have a machine learning system that produces somewhat accurate representations of user personas and stories.
For those who may not know about segmentation, Gretchen Gavett explains it here:
Segmenting, at its most basic, is the separation of a group of customers with different needs into subgroups of customers with similar needs and preferences. By doing this, a company can better tailor and target its products and services to meet each segment’s needs. We see many, many companies saying, ‘I want to get more consumer-driven and customer-facing. But sometimes the organizations don’t know how to start. I’d say you really start with a basic understanding of your consumers or customers, right? And that’s segmentation, segment by needs or behaviors.
In order to train these machine learning systems, open, global — and growing — datasets are needed. We do have some of this data. Things like publicly available information, social media, and segmentation data from brands.
The system can provide strategists and researchers with a representation of their users (i.e. it produces a composite result of Caleb, who is interested in an electric car). When further research is conducted with users, they can input that information back into the system to get a clearer picture of who it is they are designing for. A designer can then make adjustments based on that data. This is not advocacy for automating the research process, because only through true human connections and insights, can we truly impact lives.
These small optimizations can free up the designer’s time to let them think about more strategic product decisions — Fabricio.
Poking Holes in the Concept
Some important questions going forward are: Who will use this?, Why is this better than the way we make personas now? What industry/market needs this? What about things like Facebook business? What about human bias and data governance laws? How can the data be reflective of the world we live in? Is everyone authentic on social media?
To answer the first two questions, I believe this could be something that speeds up workflows for UX researchers, strategists, and data scientists.
Machine learning is also a great tool for pattern recognition, from the data fed into the model we can glean very clear insights and make informed decisions moving forward with designs. How can something like this amplify the skills of a researcher? Josh Lovejoy explains it clearly here:
Imagine a teacher compiling a reading list for her students. She wants them to grasp a certain concept which is expressed in different ways in each of the books. Simply memorizing the various answers won’t result in any practical knowledge. Instead, she expects her students to discover themes and patterns on their own, so they can be applied more broadly in the future. The majority of machine learning starts much the same way, by collecting and annotating examples of relevant real-world content (this is called“training data”). Those examples are then fed into a model that’s designed to figure out which details are most salient to the prediction task it’s been assigned. As a result of this process, predictions can be made about things the model has never seen before, and those predictions can be used to sort, filter, rank, and even generate content — Fair is not the Default.
This tool could also benefit startups trying to find the right market fit for a new product launch — it could help them test and iterate faster, know if the product is the right one or not.
I don’t have the answer to all the questions yet, as this is unexplored territory but I’m sure the answers will present themselves as I continue my research.
Benefits
More inclusive design
Earlier in the article, I mentioned that the system could help us think about use cases we may not be exposed to, and approach new users having some kind of concrete knowledge about things that matter to them. (two use cases could be designing treatments for burn victims, mobility in rural areas)
I also mentioned the system could be a tool used by data scientists. The AI/ML field is mostly dominated by data scientist and algorithm engineers and this system could lead to more collaborative work between designers and data scientists.
“Today, the best designs aren’t coming from a single designer who somehow produces an amazing solution. The best designs are coming from teams that work together as a unit, marching towards a commonly held vision, and always building a new understanding of the problem.” — Jared Spool.
Imagine this collaboration with AI in the mix.
More personalization in the user experience usually means more relevance for users, which leads to better conversion rates.
Closing
These are just my musings at the moment. In the coming weeks I hope to build out a proof of concept for this.
Thanks for taking the time to read this, any and all feedback are welcome!
