It’s Time for a Design Evolution in Machine Learning

Why the Best is Yet to Come

Let’s play a mind game: 
Select a technology that you love and go back to a time where it was brand new. Even if you haven’t been born back then or were too little too remember.
How did it look? What impressions did it make? What was the public opinion about it? What connotations did the word have? 
Who used it?

Chances are, it was considered pretty complicated, technical and a “thing for the future”. Others even feared it, since it fundamentally affected their daily lives and routines. Some already used it, but for a special uncommon purpose.
Only those with the knowledge about it rambled about all the new opportunities and saw the big picture.

That’s a stage every technology is going through when emerging.

That’s the current state of Machine Learning.

What is Machine Learning?

Machine Learning itself is a subfield of Artificial Intelligence that is skyrocketing at the moment thanks to the power of computers these days.

To put it in a simple way, it’s the computers ability to autonomously learn from data and thereby decide and predict. This can take on different manifestations in the end. Prominent examples are the Netflix recommendation system or the image scanning method OCR (Optical Character Recognition).

The History of the Smartphone

Before I focus on Machine Learning, let’s take everybodys darling, the smartphone, and replay this whole scenario with it.

Leading smartphones around 2005 where as small as possible out of the technological aspect of utility. Their only „smart“ applications were supporting or substituting computer applications as a calendar, address book or which were mainly work-related. Even foldable cell phones which came with a camera, had limited personal use, due to the complicated process of transferring images and the low resolution of them.

“While enthusiasts and non-enterprise users had found other uses for their smartphones than email and work, the main function of such devices was keeping employees within reach and connected away from the desk. Smartphones were primarily used for correspondence and light Web browsing on the train.”

Pocketnow.com

As you might know Apple came around and brought the iPhone to market which would become a milestone in the history of the smartphone and mobile computing. After time other companies jumped on the train, resulting in an incredible design evolution lasting for several years.

One reason for Apple’s success is that the product creation was lead by designers. They took into account all the things that people didn’t like about the other phones, as the tiny buttons which guaranteed for a clunky and difficult way of input. Designers saw the need for a larger screen in order to harness the full power of images. They recognized that users saw their mobile phone as a personal item and attached emotions to it. Therefore they’ve put an emphasis on aesthetic design and high quality materials, so their product would be something to “show off”.

I would go as far as to say that Apple wouldn’t have used a touchscreen in the iPhone if their had been a been a better solution/technology at this time.

I say this, because their approach was entirely based on the what the user needed, instead of what was the trend or common practice.

The Current State of Machine Learning

In the beginning I stated that Machine Learning is at the same state as the mobile phone in the years before the iPhone came.
How did it come to that conclusion?
If you compare Machine Learning and the smartphone, you can see some striking patterns in their history.

  1. The technology behind the smartphone (e.g. the touchscreen) had been around for a decade and was used in different contexts. This gave Apple the perfect basis to act upon.
    Machine Learning as a technology is already widely applied at big companies and exists as a term since 1959 when Arthur Samuel first defined it. Gartner’s 2016 Hype Cycle for Emerging Technologies says that it will take two to five years for it to reach mainstream adoption.
  2. Before the design evolution, smartphones’ creation was lead by developers focusing on performance, technical details and utility. At the moment the hype in Machine Learning is about getting incredible results, trying to replace humans (e.g. autonomous car), not about designing in coherence with machine learning in order to meet peoples needs and wishes.
  3. Few people saw the potential for disrupting every day life in mobile phones as Apple did. Today many people don’t understand the possibilities of Machine Learning and areas it can impact. Most see it as a mysterious part of AI that is used in robotical systems like Siri or GoogleCar. Though it is widely used, Machine Learning is mainly applied for research or commercial use and in products for Joe Average.

Furthermore there is evidence that Machine Learning would not only flourish by a Design Evolution, but desperately needs one.
Microsoft’s infamous chatbot Tay or “The First International Beauty Contest Judged by Artificial Intelligence” which ended up being racistic, show that ethical considerations need to be put in focus when developing artificial intelligence.
Partly due to these examples and science-fiction movies

I know that Machine Learning touches peoples lives less obviously than a smartphone and is a technology in contrast to a physical product. So comparing these two appears to be unfair right?
I am certain Machine Learning has the potential to shape the interaction between humans and machines to the same degree (or more) than the smartphone.

What is a Design Evolution?

Design Evolution is a term, coined by myself, describing the increase quality of products by putting into the focus of a user-centered design process. Crucial for a measuring a design evolution are the three following aspects:

  1. Increased Quality: the product should differentiate itself by design in a way that makes it more valuable for users
  2. In the Focus: The technology/technique/material should not be treated as feature or add-on, but seen as an integral component to achieving the best solution
  3. User-Centered: The product should fulfill real users need and respect boundaries and should not be a experiment

Adapting this onto Machine Learning means a Design Evolution would look like the following:

  1. Higher quality implementations Machine Learning concerning not only the technical side, but even more the ethical side and the integration into lives of people to provide real value
  2. Products and services using Machine Learning as a central component in their experience to solve users problems. This could be done in a dedicated approach like Anticipatory Design or by raising awareness for the technology inside the team
  3. Machine Learning applications that user-centric and were designed “user-first” manner so people really benefit from the technology

Why is the Design Evolution still missing?

Designers have a pretty complicated relationship with Machine Learning at the moment. A study conducted describes the situation pretty clearly:

“Some user experience (UX) commentators speculate that ML is the new UX, and highlight ML algorithms as being “where the action is”. […] We should now expect UX designers to regularly instigate innovative products and services.

It has been our perception that UX Design teams are simply “putting lipstick on the pig”

So why is it that designers, creative and innovative by profession, merely treat Machine Learning as a cool feature? Why aren’t they considering it as a legitim and stand-alone design material and putting it into the focus of their practice when useful?

The same study goes further and investigates on the reasons for this:

  • First of all they have found that most designers do not know the capabilities and limitations of Machine Learning and did not learn about the technology in their education. Additionally the statistical intelligence behind Machine Learning is very different to human common sense intelligence and thus seems unpredictable from time to time. This makes it difficult to envision new ways to integrate Machine Learning in products and increases the fear of potential “creepiness” of the outcome.
  • Secondly Machine Learning brings a lot of challenges for the design process, due to its huge dependence on data. Not only does this impede the prototyping of such applications in early stages, but also makes it hard to visualize and draft them. State-of-the-art tools are not prepared for this and thus do not offer the features necessary.
  • Lastly, designers consider the result of Machine Learning as not predictable and therefore hesitate to interact with the technology. The infamous examples I mentioned above played their role as well, so that nobody wants to gamble and create something „creepy“. The root of the problem of this is the missing knowledge, experience and tools to ensure that the products of Machine Learning will not hurt common ethics.

So what now?

In order to combat these obstacles, we need

  • Better education on the principles of Machine Learning
  • Better tools to visualize and prototype applications with Machine Learning
  • More research and experiences on ethical correct Machine Learning
  • Experiences in fully embedding Machine Learning into products and design processes

If you’re a designer, the best you can do right now is to:

  1. Get as much knowledge about ML as you can
  2. Expose yourself to the different kind of intellect
  3. Keep being creative and envision new ways to embed and apply ML

As I detailed above Machine Learning needs a design evolution. It’s now up to us designers, researchers, product managers, as a part of the UX community to turn it into reality. So please spread the word and do your best!

Author: Florian Oefner

Collect by: uxfree.com

Top