5 Key takeaways from UX London — Part 2
In part 1 of this article I wrote about the Design Sprint workshop by Clearleft and Jared Spool’s talk on how to make design an embedded part of your company’s DNA.
In this post I will write about a way to easily reform users’ behaviour and about overconfident machines.
3. Nuggets can help reforming users’ behaviour
On the second day of the conference “UX London” I attended a workshop about Behavioural Design, held by Jerome Ribot & Roxy Borowska from Coglode. Coglode deals with behavioural research and tells product and design teams how to use “Nuggets” to make better products.
Introduction to Nuggets
What are Nuggets and how can you use them?
Nuggets are cards with summarised research information about behavioural principles. They are written in a way that’s easy to understand and enable you to apply the research to your project.
On the front side of the card there is a category like branding, a number, the name of the behavioural principle, a short description about it, a graphic that helps you remember the content visually, a longer description of the findings and the academic paper and related Nuggets. On the backside there is a study result concerning this behavioural research and key takeaways.
In the workshop we got different tasks we should solve by applying the Nuggets. One exercise was “Imagine there is a blank paper hanging above or next to the printer. Use this paper to make people print less”. By applying some of the behavioural principles different solutions were created. While some appealed to the users’ sense of guilt, others were encouraging and some even funny. Many groups included illustrations to support their textual statement or to add an entertaining note. This along with other exercises showed how easily you can apply research and create impact with your work. It also showedhow manifold the solutions can be since there are so many ways you can have influence on a users’ behaviour.
We got some nuggets as an example. When I showed them in denkwerk, my colleagues were really interested and enthusiastic. I look forward to trying them out soon in our next projects.
4. Working with AI — Be expressive about what you are certain of and what you are not
The third day dealt with future topics like Artificial Intelligence, Anticipatory Design, Chatbots and Voice. There were great talks and in the afternoon I attended the third workshop of the conference.
I chose the workshop “Designing for AI’s Strengths and Weaknesses” by Josh Clark, the founder of Big Medium, a design studio based in New York.
In his talk and workshop he dealt with AI as a new design material and reflected on AI’s strengths and weaknesses.
These are the 3 key points:
1. AI is a design material
Machine learning is just like HTML — it’s a design material we as designers can use to create awesome solutions, says Josh Clark. It is time for us as designers to really get engaged with this design material.
So how can we use it in the optimal way?
AI can already do a lot. It can translate sketches into digital UI, it can give recommendations, and cluster and classify things. But the real work of recognising a problem and thinking of solutions is still something people need to do.
Josh’s advice: Let machines take over the things that we hate, so we can focus on what we do best.
2. Make the level of confidence transparent
When you work with AI, you realise one thing very quickly: Machines have a clear overconfidence problem. For example, looking at the following picture, what do you see?
The AI describes the picture as follows: “A dinosaur on top of a surfboard”.
The dinosaur was well recognized, but the surfboard? How can a linear scale become a surfboard? More importantly, how can the machine claim with certainty to see a surfboard even if the statement is false?
This example shows why we should treat results as signals not facts. Our job as Experience Designers is to make it transparent that machines are not always 100% accurate. Communicating confidence levels or giving vague statements rather than detailed but potentially wrong statements effects the acceptance of the machine’s answer.
It is better to let the machine give rather vague statements than incorrect ones. “I don’t know” is better than a wrong answer.
Be expressive about what you are certain of and what you are not.
An example is the following image, which was tweeted by the New York Times.
The tweet text reads “Why isn’t Italy kinder to gays?”. The computer-generated caption says, “I’m not really confident, but I think it’s a person holding a tennis racquet and she seems neutral face.” As you can see it is actually two copies of the statue of David photoshopped together so that one David has his arm around the other as they look at each other.
How you could communicate confidence levels:
This is just one example of a machines’ overconfidence, there are many more of them on the internet. As an Experience Designer you should always try to raise people’s awareness that machines are not always 100% accurate. We have to understand AI’s strengths and weaknesses and make them transparent to the user.
3. Test your data set by using different services
At the end of the workshop we had the opportunity to try out Microsoft Azure. We tested which image content AI understands and which caused problems. Most of the images were interpreted very well. Often even really complex content was described correctly. A picture Azure struggled with was an abstract painting from Jackson Pollock. It identified “a colourful kite in front of a tree” with a confidence level of 50 %.
Microsoft Azure is just one of the services which deals with AI. If you are developing a product using AI, always test your data set using different services, since every service has different strengths and weaknesses. Test your data using Amazon AWS, Microsoft Azure, Google Cloud AI, IBM Watson and more, to make sure you have identified possible weaknesses and can release an excellent product.
—
There is another key takeaway I’d like to share with you. It was from the first talk on Day 1 and it provided good advice especially for those starting new to a job.
5. Tell the whole story
This advice came from Jane Austin, Director of Product Design at the healthcare provider Babylon Health.
Jane Austin talked about a project that many designers find difficult to design and to promote: Oneself.
How do you apply for a UX job? That’s the question many newcomers ask themselves every year. Even if you are a more experienced designer and change jobs, you always have to think about how to position yourself as a good “product” on the market. What is your value? How can you convey your value well enough to get your dream job? What if your last project ended badly? Can you still put it in your portfolio?
Her advice for applying with a UX portfolio at Babylon Health is something we at denkwerk agree with: “Tell a story. Do not only show the outcome, but show the process, challenges and problems you were facing and the solution you found“.
And yes, even if a project ended differently than expected and there were problems and mistakes during product development, you can still tell this story. Show that you reflected on the process, have identified what went well and what went badly and tell people what you learned from this experience. In this way your counterpart learns much more about your personality and your way of working than by a portfolio, which only contains finished work without a description of how you came to a great ending. Tell the whole story.
Conclusion
It was a great conference with great talks, workshops and people. Thanks, denkwerk, for sending me to UX London! I’m already looking forward to next year’s lineup.
