Blog: Experience in designing for AI
Experience in designing for AI
During the course which is advanced designing for artificial intelligence, I gained many valuable experience and knowledge around AI design.
My group is aiming at designing an experience for urban farming by using AI technology. Urban farming is a community-based farming place where people can join them to grow plants in the city. It focuses more on educating people with more sustainable practices for farming and less commercial based.
We approached the problem by conducting research to understand target users journey and their painpoints. And further, we can generate insights based on our findings. Our insights and empathy that we built with people after this process lead us to brainstorm the best practices that we can implement with AI technology. This pre-deep research helped us to complete the AI design thinking exercise.
Ai design exercises and human-to-machine communication model helped us form a deeper understanding of how our concepts and ideas fall into the problems that we are focusing on solving. The very practical aspects of the exercise are that it helps us to rethink our target user and user’s journey from a broader and AI perspective. For example, for the users, instead of focusing on the stakeholders such as farm managers, volunteers, the exercises broaden our view to look into accessibility aspects for extreme users(outliner users) such as professionals in plants, people who are disabled, people who can’t not work under the sun for a long time. With a more thorough consideration from their perspective, it helps us form a more comprehensive view of how to use our product in a way that can meet more target users needs and reach the requirement of accessibility.
Secondly, the AI design exercises opened our eyes in consideration bais. When designing for AI product, it is inevitable for us to take bais into account since we are collecting various data and generating solutions through AI machine learning approach. Bais is what I never thought of from any of my previous design projects. At first, I didn’t think of any bais in the farming process. However, when diving deeper into this project and research more, I find out that even its for farming which is an objective data generating process, it involves people in it where the bais is mostly possible generated. For harvest distribution, the bias can happen from the manager’s side, which how to decide the unbiased place for even distribution is difficult. And also for the plants, some might be easily planted in certain areas while because of people’s cultural difference and geographic difference, there’s bias in which plant should be planted here.
Thirdly, by understanding the backend knowledge of how does the machine learning model work, I was able to consider different data that was needed for our product. By having a holistic overview and also a detailed consideration about what data do we need ( eg. weather data, climate data, soil condition, plant information, plant disease data, etc.) to train the machine learning model in order to build a product that is more likely to perform in its best practice to help users with the problems that they’re facing during the urban farming process.
In this whole exercise process, it helped me learn a lot and opened my eyes of what needs to be considered while designing for AI. But this exercise was conducted after sometime when we’ve already finished our research. When I was completing the exercise, it required me to try to think back and remember all those research findings and insights we got before. It would be better if we could implement this approach right after our research phase and build more informed information when completing the exercise. Besides, what I would change is to learn deeper about how AI (machine learning) works by collecting data and creating models for deep learning before our research on the technology part so that we can have a clearer notion of how it works and have a focus when we are researching.
For future users in urban farming, it is nice to gain their feedback on our farmvision product and we’d like to also focus on both sides of the users which is to consider the farm managers’ needs.