Applying Design Thinking to Artificial Intelligence. Why Should You Use It in Your AI-Based Projects?
AI product managers and designers use the Design Thinking methodology to build AI systems that are more human-centered and agile in development.
Join the DZone community and get the full member experience.Join For Free
Choosing the right project management methodology can be crucial for your project development. It will help you avoid mistakes, speed up the whole process, and support in discovering the problems of your target groups. The last issue is fundamental. Only after a deep understanding of the needs of your target group will you be able to develop a solution that will solve their problems. There are many approaches to project management focusing on discovering problems, and design thinking is one of them.
AI is becoming a more significant and critical part of our lives. AI-based products and services are everywhere, from self-driving cars to voice assistants like Siri or Alexa. AI Design Thinking is the process of designing AI systems that can operate in an unpredictable environment with limited resources in a lean, iterative way. Designing for AI requires different skills than designing for other types of technology because AI doesn’t follow predictable rules and behaviors. Today, we would like to tell you more about how it can influence your process and how you can implement it into your artificial intelligence project.
What Is Design Thinking?
Design thinking is one of the oldest (but still modern) approaches to creating the perfect development process. This approach starts with the user and puts him in the center of the whole development. His needs, emotions, feelings, and problems should be the most important things for the development team.
Explorers started to formulate their first ideas around design thinking in 60’; you can find them (for example) in the book of L. Bruce Arche, “Experiences in Visual Thinking.” Their goal was to use the tools and best practices for creative people – like painters, writers, or designers – in utility products or services development.
For some time, design thinking in the IT world was a bit forgotten, or let’s say honestly – it wasn’t the most popular methodology in the times of Agile or Scrum, but its popularity is constantly growing as implementing design thinking is given some visible improvements to your project – like faster and better decision making, helps you to get a clear vision about problems of your target group, reduce the risk of the whole project, etc.
How Does the Design Thinking Methodology Fit the Development of Artificial Intelligence Projects?
One of the challenges for design thinking in artificial intelligence is no universal approach. In Nexocode, after a few years of working on similar projects, we developed our one – mature and battle-tested process that uses a lot from the design thinking framework and solutions like super popular within the software development community Design Sprint framework. We mixed and matched our knowledge and experience to create a roadmap for every new business that wants to innovate with machine learning. It starts with AI Design Sprint workshops that are tailor-made for every client and focuses on researching AI opportunities, prototyping, and testing possible AI implementations. We believe that every organization must develop a useful artificial intelligence project to understand why, where, and how they should develop it, and that’s why our AI Design Sprint is focused on those topics. It is just the beginning, but every next step we take once our client decides to move on with the project is iterative by design.
When implementing design thinking into AI development, the team is the key. Therefore, it is essential to have a team of experienced AI experts. They will play a significant role in the whole process, and their knowledge will influence the project.
Why Machine Learning Projects Need a Human-Centric Approach?
Designing AI requires different skills than designing for other types of technology because AI doesn’t follow predictable rules and behaviors. It means that there is a need to create human-centric solutions as much as possible, considering the needs, emotions, feelings, and thoughts of people who will use those technologies daily while considering all the problems they may face while using such AI-based products or services. The feasibility of designed solutions and their implications is not so apparent as in normal software development. Machine learning projects need good, ethical design and solid data sources. Each project is different, but project managers’ data science knowledge is crucial for successful research and development.
Designers should focus on AI design thinking to create human-centric AI products and services. That’s why it is essential for AI designers to follow the same process of design thinking as other types of technology but also consider the emotions, feelings, and thoughts of people who will use those technologies daily while taking into account all problems, including AI ethics, they may face while using such AI solutions.
In AI projects, accountability is vital as AI-based products and services are already influencing our lives daily.
Designers conducting AI design thinking should consider all possible scenarios when using AI in various aspects of people’s lives while considering different types of risks during actual usage. For example, who should be held accountable if an AI system makes a particular decision? Our AI system's decisions final, or is there human supervision?
Deep learning systems usually work like black boxes. Their decision-making process is not explainable in similar ways as we might take a decision. To some extent, all AI solutions can and should be explainable. However, AI designers need to understand that AI is not a magic box, and there are some rules on how it works, which means people may know why AI acted in such a way during specific scenarios.
AI-based products and services might not be easy to trust. AI algorithms are often opaque, and a lack of explanatory AI can lead to over-reliance on AI. Design thinking is the tool that allows you to build trust in AI by designing systems that provide clear feedback loops for users so they understand what an AI algorithm does.
Human-AI interaction is something new that has to be treated differently than standard Human-Computer interaction. There are several Human-AI interactions best practices and recommendations. Design thinking methodology is a great framework for AI-based products and services because it encourages you to think about AI from an end user’s perspective and focus on possible interactions.
The main advantage of design thinking over other methodologies, in this case, is that it allows designing AI solutions by considering input data, algorithm process, output, and all possible scenarios where AI can be used. That way, designers have more control over AI decision-making processes, making AI much less ambiguous than programming languages.
Stages of Design Thinking in Artificial Intelligence
One of the most important things to understand when talking about design thinking is stages. Design thinking is a simple process where the next stage emerges from the previous one and can be started only when the previous step is finished.
This stage concentrates on feeling empathy with the users of your product. You should gather many people representing different societies, mindsets, experiences, and groups and work with them to discover what they feel, think, and expect. Think about how you can improve their lives with your product or service. Once again – remember that at the center of your development process is always a human and his needs. You’re implementing a back-office AI platform that helps your manufacturing process, and you think there is no one to empathize with? You couldn’t be more wrong. There are plenty of stakeholders involved in every process, and this first phase is about feeling and future goals and opportunities. This phase became more complicated when implementing artificial intelligence because you might need to know something about ML models, neural networks, or data analytics. It is essential to think about the feasibility of artificial intelligence at the beginning of the project to avoid the complicated implementation process into the existing solution.
After interactions with people representing different worlds, you can define your target group and target challenge. Think about AI opportunities. Choose a group with specified and justified needs and concentrate on their problems to implement your artificial intelligence solution. You will see the whole picture clearly and choose the one you want to address with your project. It’s the moment to ask questions and search for insights and go deep into the problem.
The next stage is about finding solutions for the problems of your target group. You gather your team, and you brainstorm all the possible ideas you have in mind. The goal of this stage is to unleash your team’s creativity and find some new and uncommon way of solving the problem of your target group. You can formulate which AI algorithms, tools, and techniques to use in your project at this stage.
After brainstorming with your team, you choose the most interesting ideas and changing them into prototypes – for example, MVPs - to collect the knowledge as fast as possible. There is no need to develop a full-scale AI solution at this stage because this is a time-consuming process. The main goal of this phase should be to have the ability to learn. With this approach to artificial intelligence development, you will choose the best one or choose the best things to develop a perfect final piece of software.
This is the last but probably one of the most important stages, as it can help you identify and remove the problems with your product. It’s the moment when you show your prototypes to the target group or test in a close-to-real environment selected in the first stages. You observe them – how they react, how they use your product, and what emotions to do they feel. Is your solution solving their problems? If they do not like it, it’s the moment to go few steps back.
Advantages of Design Thinking
Satisfy Stakeholder’s Needs
Increased customer satisfaction and business adoption of a company’s inside software is one of the most significant advantages of implementing design thinking in every project (also inside artificial intelligence-based ones). Users of the products build with design thinking methodology declare higher satisfaction when they use them. As your user is always at the center of product development, customer satisfaction should always be your main goal.
Increased ROI of AI Investment
Rest assured that the time you spent on design thinking exercises secures your long-term AI investment. Every business has its characteristics and needs. That’s why the implementation of machine learning should be tailor-made. Design thinking helps find pain points and define a business case for AI at your organization and therefore aids in turning your AI dreams into a profitable investment. You can read more about this in our ROI of AI projects article.
The ideation phase of the design thinking process is designed to think in a non-standard way, to think out of the box. This approach can lead you to ways of solving problems that no one ever considered. This might help you create a very innovative solution and stand out from the crowd on the market – or even become a leader in it!
Reduced Risk of Failure
When you investigate your target group and their problems in every detail, you will develop a product that will satisfy their needs increases, and the chance that your product will be successful is higher.
Design thinking methodology is going to stay with us for a long time. This proves that it can be used for effective and useful approaches to artificial intelligence products or services development.
But it’s not an easy task, and specific organizations might have problems. Therefore, it is essential to find a reliable partner who will support your team during the process - starting with the strategic meetings and ending with the successful testing and development phase.
Published at DZone with permission of Dorota Owczarek. See the original article here.
Opinions expressed by DZone contributors are their own.