The intelligence augmentation design toolkit and the associated workshop demystifies machine learning and helps non-tech experts to create smart service concepts. Using the toolkit requires neither a technical background nor any coding skills. The toolkit consists of four sets of cards, a map, two canvases and a booklet.
- What was the challenge/brief? How to teach non-tech experts to design future smart concepts.
- What was the context & industry sector? Service Design & Machine Learning
- What was the challenge/brief? How to teach non-tech experts designing future smart concepts.
- Objectives: To bridge design and tech, make the complex simple, stay tangible.
- Outcomes: IA design toolkit and associated workshop. The toolkit consists of four sets of cards: touchpoints/channels, ML interactions, customer segments (not available with CC license) and unexpected bug cards, a shopping mall map, two canvases: confusion matrix and smart service canvas and a booklet containing a summary of the workshop lectures and a dictionary.
- Target market: Non-tech experts.
What type of research did you do and why? This project was created to solve a problem we had found ourselves. The authors of the toolkit felt that there was a need to approach machine learning with a fun and easy-to-understand way. We felt we needed new tools to the concept without writing code and also to bridge the gap between service design and data science. Due to the own experience and being the target group ourselves, the research part was rather light and most of the time was spent on testing and iterating the toolkit in practice.
For the background research, we used netnography, benchmarking: We wanted to find out what tools existed for designers to learn machine learning in a tangible easy way. We concluded that by that time, there were no tools specific to machine learning and service
design. We also benchmarked design games for mapping out different approaches in the field. To solve the problem, we mapped out the needs and hopes of our design and data science teams. We used online chat tools to gather information from different designers: how they felt about the topic and what kind of questions they had in mind.
How many participants were involved in each stage of research? Benchmarking was done by three people: service designer, data scientist, and tech wizard. Visual designer joined when making the first drafts. Testing of the first ideas was done with 25 people from the in-house design team and 4 people from the in-house tech team. We consulted a team of 30 designers.
Explain how you gathered customer insights
Non-formal discussions with several designers and tech people from in-house design and tech teams of our company. Own experience - the insight from the core team. Testing the initial concept and drafts with experts from in-house design and tech team.
Describe the specific design tools, methods, and processes
Prototyping and contextual group interview: When the demo version of the toolkit and first draft of the design was created, the script and the first version of the tool was shown to a small test group.
Why were these specific tools and methods used?
We work with lean style: build-measure-learn.
Outline key insights and how you used research findings to move the project forward
Non-tech experts felt it was easy to approach a complex topic, when using tangible paper tools. Cards helped with categorising different functions in a more memorable way. The illustrations made difficult topics easier to remember and more likable. For example explaining machine learning bias in the form of an illustrated bug, who comes to visit was considered humorous. The canvases helped mapping out complex topics such as a confusion matrix in an easy and fun way which was still useful from data science point of view: the terms remained the same.
Clarify how you developed and tested the product or service concepts
Core-team created the first concept and prototype for the kit. Visual designer joined the team to create the visual look and feel for the package. The first version was tested with 5 experts. Iterations and prototyping continued: Based on the feedback of the test group, we iterated the concept and materials and run the workshop for a bigger test group of 25 designers. Designers worked in 3 teams, after running the workshop, we had a separate feedback session for improvement ideas.
Based on the feedback of the bigger test group, we ran the workhop in public event for 50 designers. People were sigining up via Eventbrite. Again we asked everyone to leave us feedback. Since then we have iterated after each workshop and for example, updated all materials and increased the page count in the booklet. Also, the lenght of the workshop has been developed into 1,5, 3 hours and a full day versions.
The associated workshop has four versions:
- 1,5, hours: basic machine learning for everyone and fast concept creation overview
- 3 hours: machine learning for everyone, concept creation, presenting concepts to teams.
- Full day: machine learning for everyone, concept creation, prototyping and presenting to teams.
- A commercial version for the customers. Tailored for company-specific needs and use cases.
The Intelligence Augmentation Design Toolkit
- Channel/touchpoint cards, 29 cards introducing different enablers
- Machine learning interaction cards: 10 cards, 4 categories explained: predict, recognize, personalize, uncover structure.
- Customer segments cards & template: 10 customer segment cards for workshops (a market research company 2017 co-operation) Free template for creating your own segment.
- Unexpected bug cards: 11 bugs explained
- The map: in the free version this is a map of the shopping mall, to be used to map out the customer journey and the machine learning spots.
- The confusion matrix canvas: A canvas to map out 4 different scenarios for machine learning event.
- The smart service canvas: A canvas to map out the concept idea and basic machine learning features
- The booklet: A short summary of the workshop lectures, dictionary and in the new version: instructions added for how to create the concept.
Benefit for the customer: Designers and non-tech experts learn machine learning in a fun and engaging way.
What were the benefits to the client?
For organizations, companies, meet-ups, and festivals, this has been a great way to get people engaged in a complex topic and innovate new concepts for the future world.
What were the effects for the organisation?
This has become a popular and known tool for our company to reach people across disciplines and engage people from different domains. We have run several events in Germany, Finland, and UK.
Stakeholder feedback: One organization gave us feedback saying this helped them to: “Increase of the knowledge of the methods and theory of machine learning as well as applying design thinking with intelligence augmentation. Participants also learned to collaborate with people of various backgrounds on complex issues as well as increased their capabilities in designing services with a machine learning/intelligence augmentation elements.”
Were there any benefits for the competition or the market as a whole?
The tool will help service designers to bridge the gap between service design and data science, which will benefit the whole design industry. Designers can make more relevant concepts for emerging tech.
Describe cause and effect related to the project
- 600+ trained non-tech experts in 3 countries: Finland, Germany, United Kingdom
- 1000 paper copies of the booklet distributed
- free online tool available with CC license.
- word-of-mouth: also external meetups and groups using the tool by themselves. (we’ve got pictures and letters)
- All international sites of our company are using the tool for external events.
- Several requests coming in for arranging workshops - a clear need for this type of work in many different industries.
Describe the scale of impact locally and internationally
- Demystifying machine learning. Instead of the somewhat misleading media hype of AI taking over the world, the tool has helped designers to see what machine learning is good at and what it is not. The tool has helped designers to focus on what are the current problems what us as designers should concentrate on solving.
- The tool has helped non-tech experts to grow in their machine learning path
- The tool has helped our company to build our own learning portfolio as well as offering
- The free version of the tool is available for everyone
- The free workshops offer an interesting program for different meetups, hackathons,
lectures and conferences.
What was the impact on the client department, wider organisation & other stakeholders?
- Clients have been so far mostly networks and events, the toolkit has offered participants an interesting content and helped them to learn in a fun and engaging way.
Did impact match initial objectives?
- Yes, the toolkit has been well received.
The journey has just begun. The project has opened us as designers of the kit a new way of thinking and experiencing design and tech. In today’s world physical and digital design are inseparable. We need tools to bridge service design and emerging tech. Instead of artificial intelligence, we use the term “intelligence augmentation”. We like this term because we think humans should not be replaced, instead, machine learning can augment our skills and help us do tasks faster, also letting us focus on more meaningful things. However, we need to be active in the process of making this happen.
The main takeaway for us as designers of the toolkit and workshops have been that machine learning is something that everyone can and should learn. In order to stay relevant in the world ruled by algorithms, it is important for designers to understand the logic behind the algorithms and what are the items we can and should have an impact on: such as the content of training data. In our materials, we cover topics such as learning loops, discoverability, training data, machine bias, designer ethics and human bias. We learn what can go wrong and how to overcome troubles with customer service strategies.
There is a still human designer designing the machine learning systems, there are still humans finding the right problems to solve. We as designers are all responsible for designing smart future services.