The world’s leading publication for data science, AI, and ML professionals.

Surrogate Models Can Help UX Designers Build Better AI Products

Why AI products are so difficult to design

Photo by DeepMind on Unsplash
Photo by DeepMind on Unsplash

User experience (UX) tools help designers create better products by making it easier to understand how users interact with them. However, there are few UX tools specifically designed for artificial intelligence (AI) products. This is a problem because AI products are becoming increasingly common and they present unique challenges for designers.

Building AI applications is hard. Not only do you need to have a good understanding of the AI algorithms themselves, but you also need to be able to build user interfaces (UIs) that allow users to interact with the AI in a meaningful way. Unfortunately, there are not many good tools out there for building good UIs for AI.

In this article, we will discuss the problem of UX tools for AI products. We will first explain why AI products are difficult to design and why existing UX tools are not well-suited to the task.

We then discuss a method called a surrogate model, which helps to better understand a typical AI "black box" that is difficult to explain and where errors are hard to predict.

Why AI products are so difficult to design?

As an AI developer, I would say that AI products are difficult to design simply because they are complex systems with many interacting parts. This complexity makes it hard to understand how users will interact with the product and to predict how the product will behave in different situations.

Traditional UX tools are not well-suited to the task of designing AI products. They are often too simplistic and do not take into account the complex interactions between the various parts of an AI system.

One of the biggest challenges with AI products is that they are often designed to be used by people who are not experts in the field. This means that the products need to be designed in a way that is simple and easy to understand.

The problem with UX tools for AI products

It’s no secret that the current state of UX tools for AI products is far from ideal. There are a number of challenges that need to be addressed in order to make the experience of using AI products more seamless and user-friendly.

One of the biggest challenges is the lack of integration between different AI products. This can make it difficult for users to move between different AI products and can lead to a fragmented user experience.

Another challenge is the lack of standardization in the way that AI products are built. This can make it difficult for users to understand how to use different AI products and can lead to a frustrating user experience.

Finally, there is a lack of tooling specifically designed for AI products. This can make it difficult for designers and developers to create AI products that are user-friendly and easy to use.

In order to address these challenges, it is important for UX designers and researchers to work closely with AI product teams. By working with AI product teams, UX designers and researchers can help ensure that AI products are user-friendly.

Surrogate model

Building better AI products is hard. There are many factors to consider, from the algorithms themselves to the data they’re trained on, to the interfaces and user experiences that let people actually use them. To design good UX for AI products, we need to understand not just how AI works, but how people interact with it.

In engineering, there is a method called "surrogate models" that can help better understand the AI behavior.

Surrogate models are mathematical models that are used to approximate a complex real-world model. In general, a surrogate model is less expensive to evaluate than the original model and can be used to optimize the design with experiments or to understand how the AI is reacting to some inputs.

Decision tree

A Decision Tree is a supervised learning algorithm that can be used to create a surrogate model. Decision trees are a type of non-parametric model, meaning that they do not make any assumptions about the underlying data distribution.

Decision trees are a popular choice for surrogate modeling because they are easy to interpret and can handle both numerical and categorical data. Decision trees can easily be trained with Scikit-learn and give you the ability to understand how the decision process is made.

Let’s have a look on how a decision tree can be trained with only a few line of codes with to Scikit-learn:

In this example, we have four data points called "X" and the given outputs "Y" to train the model. A data point has four dimensions (e.g. [0, 0, 1, 1]) and each dimension is binary–0 for no and 1 for yes.

One dimension could be "is the temperature < 20" another could be "is there people at home?", another "is the fireplace on fire?" and so on and so forth.

In this simple example, we may have a complex AI system that decide when to use the heating system based on several parameters and discover–after training the surrogate model–that only two variables are important (temperature and whether someone is at home).

In other words, a surrogate model is a simpler model that is trained to mimic the behavior of a more complex AI system. The idea is to gain insights into the from a complex AI system by understanding the behavior of the surrogate model.

Conclusion

In the past few years, there has been an explosion of new AI products and services. This is largely due to the advancements in machine learning and artificial intelligence technologies. As a result, there is a growing need for UX designers who can create user-friendly interfaces for these products.

However, designing for AI presents a unique challenge. UX designers need to account for the fact that AI products are often powered by complex algorithms that are constantly learning and evolving. This means that the user interface needs to be flexible enough to accommodate these changes.

As AI is increasingly being used to create and enhance products, the need for UX tools that can help design and test AI-powered products is also growing. However, the current state of UX tools for AI products is still in its early stages, and there are few tools available that are specifically designed for AI products.

This lack of specialized tools is a major challenge for UX designers who are working on AI products, as they often have to rely on general-purpose tools that were not designed with AI products in mind.


Curious to learn more about Anthony’s work and projects? Follow him on Medium, and LinkedIn.


Related Articles