Ideation and Design Thinking for an Artificially Intelligent Product
Design thinking is a process of finding innovative solutions to solving problems. It transforms the way we develop the products, strategy, and processes that are used in our day-to-day activities. This is a human-centered approach to what is technologically feasible and economically viable.
Overview of a Use Case: Combating Distracted-Driver Behavior
In this fast-paced world, multitasking has become a common way to save time—even while driving. Many road accidents take place where people lose their lives due to the distracted behavior of drivers. People behind the wheel need to avoid unsafe behaviors, such as talking to co-passengers, using mobile phones, and eating.
To combat the issue of distracted driving and minimize the number of accidents, the plan is to investigate how artificial intelligence (AI) could automatically detect the distracted activity of the drivers and, based on the type of distraction, alert them as a safety precaution. We envision this type of product being embedded in cars.
This use case explicates how a cross-functional team first delineates the problem and its parts and then how a developer solves this problem using AI.
This five-part series of articles explain the entire process of how such an AI-based product would be created from scratch, taking you through the steps of conceptualization, research and development, and finally productization. While we give a sense of the entire process here, we take advantage of Intel® AI technology to emphasize the AI research and software-development part of the process.
Applying the Process
Conceptualizing a Product with a Cross-Functional Team
Forming a cross-functional team is quintessential to rapidly building a complex product or solution.
Most AI products are complex and require significant financial, legal, deployment, and design considerations. This complexity makes it important to have a cross-functional team that enables quick ideation and decision making. The team needs to come together with a big hairy audacious goal (BHAG) attitude so as to promote the team’s initiative rather than an individual member. Enable initiatives carved by the team, rather than slice them to the bare minimum.
Figure 1. Cross-functional team
During conceptualization, the cross-functional team generally has decision-making representatives from Sales, Marketing, Product Design, Finance, Legal, and Research and Development, as illustrated above. Commitment to the goals of the project is important for this team to conceptualize a viable product.
The project sponsor (generally senior management) requests the allocation of the team along with signed goals that are committed to by the leaders of all the major divisions/departments in the firm, as illustrated above. The leader then assigns someone to work on this project full-time.
The development team requires experience with the following:
- Python* programming, including basic packages in Python (such as NumPy, scikit-learn, and OpenCV)
- Data structures, such as list and dictionary
- Image processing, machine learning, deep learning, and neural networks
- Web-application development
- Python* web frameworks, such as Django*
- Video processing
- Experience in deep-learning frameworks, such as TensorFlow* and Keras*
- Intel® Movidius™ Neural Compute Stick
A basic understanding of client-server architecture is an added advantage.
The problem that we are trying to solve—changing distracted-driving behavior—is of interest to the following industries:
- Automotive Original Equipment Manufacturers (OEMs) (for example, Valeo* and Visteon*) who want to create products that improve driver safety by using the latest technology to alert the driver not to be distracted.
- Insurance companies (for example, Liberty Mutual, Geico) who want to improve driver safety, thus preventing costly accidents.
Using the Five Stages of Design Thinking
To solve complex problems, the design-thinking process, which alternates between divergent and convergent modes, has five stages: discovery, interpretation, ideation, experimentation, and evolution.
Stage 1 Discovery
In this stage, we try to empathize with the problems and needs of drivers. Everyone in the team has to understand, observe, listen, and manage customer expectations, experiences, and motivations. The information gathered in this stage will be used in the next stage to define the development of the product.
During this stage, we collected a set of characteristics for the behavior-enhancement system:
- It should not be a distraction by itself. (Remember the radar detectors of yesteryear.) The system should provide the driver with minimal subtle alerts based on real distraction conditions.
- It should be a learning system. If the driver has ignored the alert or tried to reset the system alerts at any point, that particular distraction should be automatically ignored for some time.
- It should be easy to use as a part of the vehicle and should not be an impediment to the driver or to passengers.
Stage 2 Interpretation
The ideas collected in the discovery stage are now converted into meaningful insights. Here, most of the questions are “what” questions. The team is engaged in storytelling based on the insights. Then, the designer team uses the storytelling to gather ideas for establishing features, functions, and other elements that allow them to solve the problems.
Here, since all the participants in the team were drivers, we discussed what the distractions were and how much they affected the driving. Statistics of how many accidents do happen because of driving were also discussed.
At the end of this stage, we identified the following features:
- Price equivalent to about 1/5 of average insurance of the car/year
- Seamless fit onto the vehicle
- On-off switch easy to use
- Subtle alerts based on announcements, beeps that are customizable by the driver, or flashing lights pointed at the driver (so as to avoid disturbing passengers)
- Ability to use the built-in systems such as the speakers, radio, screens, mirrors, air-vents
- Ability to factor in car movements (For example, if the car is at a complete stop, a slight distraction might not trigger an alert.)
Stage 3 Ideation
The goal of this process is to develop fresh ideas from the diverse group of people. This process lasts for an hour, during which people generate ideas without any constraints. Many ideation techniques—such as brainstorming, brain writing, worst possible idea, and SCAMPER—are used to identify possible solutions to the problem. After the ideas are generated, the team selects a few promising and exciting ideas to use in the final product.
We selected the following ideas:
- Mobile aids to ping drivers when they become distracted
- Devices in the car, such as cameras and flashing lights, to aid with curbing distraction
- Steering-wheel accessories to buzz the user
- Reporting aids to provide a distraction score at the end of the ride
- Windshield devices to project distraction dangers
- Seat stimulations to curb distraction
- Equipment to play warnings on the car speakers
All the above ideas were collected and put together in a document to be addressed in the next stage for feasibility.
Stage 4 Experimentation
At this stage, the design team debates and critiques all the solutions provided and selects the one that has minimum viable product and feasibility. Prototypes are shared and tested within and outside the team. Based on the user experiences, the prototypes are investigated and either accepted, improved, re-examined, or rejected. At the end of this phase, the design team has a better idea of how drivers think, feel, and behave while interacting with the end product.
Stage 5 Evolution
During this iterative process, the best solutions identified during the experimentation phase are rigorously tested by the designers and drivers. During testing, people give feedback on how they feel, what they think, and how they understand the product. Many improvements and alterations are made to meet and exceed people’s goals.
Formulating the Final Concept
The final concept narrative looked something like this:
- The product could be a camera with a chip that can detect a distracted driver.
- The product could connect to the mobile phone for configuration and locking, allowing undistracted driving.
- The product would be easily installed by a driver or be embedded into car systems.
- The product would be the best possible assistance to a driver and not a distraction in itself.
- The product would learn from driver behavior and from the environment so as to automatically adapt its behavior.
This narrative, in addition to other idea charts and illustrations voted to be the best, would be then stored as a final concept.
Figure 2. Idea chart of final concept
Handing off Assets to the Development Team
The development team is given the following assets:
- The final concept narrative
- Charts, pictures, and other artifacts used and saved by the team
- Ideas discussed during ideation
- Features collected during interpretation
The second article of this Combating Distracted-Driver Behavior series, Experimental Design and Data Preparation for a Distracted-Driver AI Project, covers how research and development helps you to build your project. It mainly discusses how to prepare a dataset, how to approach a solution, and how to create a topology and design for an experiment.
Part 1: Overview of a Use Case: Combating Distracted-Driver Behavior
Part 2: Experimental Design and Data Preparation for a Distracted-Driver AI Project
Part 3: Training and Evaluation of a Distracted-Driver AI Model
Part 4: Designing and Fine-tuning a Distracted-Driver AI Model
Part 5: Overview of Productization for This AI Project
Join the Intel® AI Developer Program
Sign up for the Intel® AI Developer Program and access essential learning materials, community, tools and technology to boost your AI development. Apply to become an Intel® Student Ambassador and share your expertise with other student data scientists and developers.
Product and Performance Information
Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.