Instead of an interview, the first episode covers what to expect from coming episodes.
Hello and welcome to Design for AI, I’m Mark Bailey. Welcome to episode 1
These podcasts will normally be about interviews and talking about how to address the design problems of machine learning but for the first episode I wanted to talk about what to expect in this podcast.
For those listening who work in this area you are familiar with the current landscape for developing ML is still in the gold-rush stage. There is no one dominant player so a lot of big companies and countries are all trying to competing for any edge possible. The main focus is on getting their app out as quickly as possible. Luckily we are still in the grace period era of new a technology where the capabilities of ML still impress people enough to overlook a whole lot of rough spots.
With all the focus on getting the product out no one is looking at how ML changes the User’s experience. Thats where I come it. So Why me? Well, no one else is. I’ve searched. The podcasts I’ve found so far are talking about the technology side, usually how to develop an AI model there are even some podcasts about the business side of ML. So I’m starting this podcast to talk about design, it is something I find super interesting and I’m surprised No one else is talking about how to make machine learning work better for people.
Well, That is what I hope to accomplish with this podcast anyway. I want to talk to experts in the field to find out how they are dealing with the design challenges of the extra hassles and taking advantage of the extra capabilities that come along with AI.
But I can’t do this alone. I’m going to need your help. Like anything else creative it is better to get started than to get it perfect, I’m an expert in designing software for AI not podcasts so on that note, just like any iterative design. I need your feedback to get better. You, yes I’m talking to you, as the listener, are part of the discussion. I need to know what you are interested in hearing about, what questions do you have? What can I do better? I need you to let me know.
To leave feedback use the voice recorder app on your phone and make sure to give your name then email it to firstname.lastname@example.org
To give a little bit of back story of what my motivation for this podcast is. Years ago when I was working for IBM research, I was really enjoying designing for accessibility because of the extra complexities required to solve some of the universal design puzzles Universal design is no joke when you are trying to design for every group of people of course the deeper you dive the more groups with more and different needs that can compound the problem
or the needs can conflict with each other. Sooner or later, there gets to be too many exception cases to juggle enter machine learning for customizing the UI. Of course, It was early so machine learning didn’t work well but it was enough to peak my interest. I had to teach myself anything user experience related for machine learning. Since all of the attention AI has been on getting the technology better and now the technology is finally getting there.
So who is the Target audience for the podcast? Who do I think will find it interesting? I want to Help the developers who get stuck developing for a ML product without any UX help I want to Help ux designers learn the AI specific problems that are not issues with normal software development cycles. I want to Help PMs know where AI projects can get derailed and what to keep an eye out for. But really anyone interested in ML should find something interesting. I will try to explain any terms I need to any time I need to dive into a more technical area. One caveat is that I use AI and ML interchangeably.
Pretty much every ML development podcast I’ve listened to explains the difference so I’ll leave it to their better explanations.
That leads us to the things we will be talking about.
For the UX designers out there
Trust, how to build it, how to lose it. How AI can help ux processes for better answers. How to improve the jobs you already do as a UX practitioner with AI. How AI affects the GUI in terms of Interaction design. What to look for in user tests to see if the AI is helping or if you are getting users adapting to the test environment. How your user personas affect which models you should build.
For any developers
Special problems AI presents to software development process. What to do when you are designing a chatbot, recommendation system . How to choose the right AI algorithm for a better user experience. How different development choices, like the number of layers, can affect the experience. How to ensure consistency of the experience by tracking data, trainings, and models.
For the PMs who are listening some of the topics are.
How to have an AI design strategy for your company. How human should you design the AI to be when people interact with it. How to design around biases in data, people, and AI models. How to optimize AI for marketing campaigns. How and why to create an ethics plan for make sure AI is improving the users experience. How to tie in UX of AI into the business value so it’s not just a flashy word.
…and most importantly for everyone
Identifying the pitfalls that need to be avoided and of course when is it OK to be lazy for the unimportant stuff and when the software needs extra attention to be designed well.
Because right now as you know there is a lot of mistrust of AI around privacy. Can you trust motivation of the app? or the company? Because It is so easy for a company to lose all customer trust, in the blink of an eye if these things are not designed right.
That leads me to my ultimate goal. It has to do with trying to solve the black box problem. For those not familiar with the black box problem. It is the idea even when all the code is known all the input data is know, all the hardware and training methods are known there is still no way to know how the AI will react in every situation. That is the black box. The AI can give unknown results at the most critical moments.
Now, I think it is possible to solve this problem.
We have the info, it’s a computer. Reduce it down and it is basically just doing a lot of math as fast as possible. This is the perfect problem to be solved by UX. It is taking an overflow of information and prioritizing it then presenting it in an understandable way Basically it is one of the worlds hardest visualization problems. If you are familiar with machine learning, you know the first two AI winters were caused by over promising what AI can do. To avoid a third AI winter the black box problem, I think, needs to be solved. Otherwise people will keep falling back to the troupe of AI from movies because there is no way to know if they can trust machine learning.
What can you do to help?
Your first lesson in ML is to learn how to help train your your podcast agent by clicking subscribe or writing a positive review.
I hope you found this first episode interesting and that there was topics I mentioned you have been wondering about too. If not, well send me feedback on what you are interested in. If you did hear topics you were interested in, let me know which ones you want to cover first.
Also since this is a podcast, if you would like me to use your voice use the voice recorder app on your phone and make sure to give your name then email it to email@example.com
ML is such a powerful tool, it only makes sense to design the AI to help people as much as possible.