Case Analysis: How to Make Judgments in Different Scenarios for AI Product Design?
-
Think independently, seek truth from facts, persevere, and compensate for shortcomings with diligence.
— A Broccoli
Tom Michell provides the definition: For a certain type of task T and performance metric P, if a computer program improves its performance on T as measured by P with experience E, then we say this computer program learns from experience E.
Typically, to define a learning problem well, we must clarify three characteristics: the type of task, the standard for measuring task improvement, and the source of experience.
Take a handwriting recognition learning problem as an example:
Deep learning is a specific type of machine learning with powerful capabilities and flexibility, representing the vast world as a nested hierarchy of concepts.
Deep learning is inspired by the structure and function of artificial neural networks in the brain. Its rise has driven the practical application of AI technologies in industries such as facial recognition.
With technological advancements, a large number of AI-based products have emerged in the market.
Product design is inherently a very challenging task.
AI products, due to the complexity of their application scenarios, make product design even more difficult. For product managers, it requires deeper thinking about real-world application scenarios when designing products.
At the beginning of The Elements of User Experience, there is the following passage:
In product design, we pay too little attention to user experience: the products we create are meant to be used in the real world. During product development, people focus more on what the product will do. User experience is often overlooked—how the product works—yet this is the very factor that determines the product's success or failure.
User experience does not refer to how a product itself works; it refers to "how the product interacts with the outside world and functions," or how people "engage" and "use" it. When people ask about a product or service, they are asking about the experience of using it. Is it difficult to use? Is it easy to learn? How does it feel to use?
One particularly interesting excerpt stands out:
During product development, people focus more on what the product will do. User experience is often overlooked—how the product works—yet this is the very factor that determines the product's success or failure.
Let me explain the meaning of this statement. There are two layers of understanding:
In product development, product managers tend to overlook the interaction between the product and real-world scenarios. This is especially true for AI product design, where usage scenarios are complex and often require the product to make judgments in different situations. If the product manager does not thoroughly consider all possible scenarios in advance and prepare design solutions, the user experience of the AI product may be very poor.
Let’s illustrate this with an example of an AI product.
Recreating the real business scenario of "pedestrian red-light violation detection":
At a crossroad, traffic lights are installed to regulate pedestrians and vehicles. When the red light is on, neither pedestrians nor vehicles are allowed to pass. When the green light is on, both pedestrians and vehicles can proceed.
However, there are always people who violate traffic rules and run red lights. In such scenarios, we can use AI to better monitor pedestrian red-light violations. We hope that if a pedestrian A runs a red light, our device can record this person and upload the violation event information—including the time, location, and the pedestrian's face—to the backend monitoring system.
The above requirement description is quite broad. Let’s think about how to break it down into algorithms (AI) and hardware implementation.
After the requirement document is finalized, it might seem like our product is about to succeed. But let’s look at this image:
1. What is this image?
As a promotional advertisement, Ms. Dong Mingzhu, printed on the body of a bus, was captured running a red light and appeared on the local pedestrian red-light violation exposure board.
2. Why did this happen?
When the product manager drafted the requirement document, they only described the need to detect pedestrians running red lights. They did not consider scenarios like advertisements on vehicles, nor did they think of solutions in advance, leading to the situation shown above.
Teacher Yu Jun mentioned in his new book Yu Jun’s Product Methodology:
Technology itself does not create value; technology must be applied to products.
No matter how advanced the technology is, if it is detached from products...
Let’s think about the real business scenario: traffic lights are installed at crossroads to maintain traffic order.
When the red light is on for Lane A, pedestrians, electric bikes, and bicycles in Lane A are not allowed to pass. At this time, vehicles in Lane B can proceed. If a vehicle in Lane B has a printed human image on its body, what might happen?
Perhaps the initial product design did not delve deep enough, overlooking such scenarios, which led to the situation where Ms. Dong Mingzhu, printed on the vehicle, was mistakenly identified as a pedestrian running a red light.
Now that we understand the real scenario, how can we do better if we were to start over?