Six personal takeaways from the City of Helsinki’s AI Experimentation Accelerator
What do local governments do to encourage innovation among their employees? The answer, at least for the City of Helsinki, was to set up an AI Experimentation Accelerator, and encourage employees to submit ideas for new uses of artificial intelligence (AI). Over the winter of 2019–20, seven projects were taken forward within the accelerator, from the 31 ideas submitted. A second wave has already started, with consideration of possible projects taking place at the moment. What are the city government’s takeaways from the experience so far?
1. City accelerator projects must be selected to fit the city’s wider priorities for digitalisation
All the projects selected had a clear focus on improving the lives of citizens or making work easier for employees. This chimes with the broader aim of all the city’s digitalisation work, which emphasises the importance of people. The purpose behind the accelerator is to support a culture change within the city government, and particularly to encourage more experimentation. The shared focus was therefore an important feature. The overall evaluation of the projects focused on novelty, innovativeness and effectiveness, and also considered the feasibility of the project.
2. City accelerator projects should provide wider learning, rather than just ‘interesting projects’
The city wanted to select ideas for experiments that would provide learning for the future, and could be scaled up to benefit the whole city. The seven projects are therefore very different, so that the city can apply the learning from each in different ways. Of the seven, four involved text analytics, one statistical analysis, one computer vision, and one a recommendation engine. These different approaches to machine learning and AI support the city’s vision of using the projects as pilots, to explore a range of different ways of using AI. For example, one project known as Löytö, or discovery aims to create a recommendation engine to help people to find suitable cultural services based on their search history. This may eventually be scaled up to provide recommendations engines for all kinds of services across the city.
3. City accelerator philosophy clear principles and ideas
The idea of the Experimentation Accelerator was to encourage innovation among employees. They were therefore all encouraged to submit ideas for short-term ‘agile experiments’ using AI. The approach is consistent with principles such as design thinking, with experimental design viewed as one method of designing something step-by-step, testing all the way. The projects use active co-creation involving employees, users and enterprises, and encourage wide communication. One of the success criteria is documenting and sharing the lessons with the rest of the organisation.
4. City accelerator was also designed to overcome specific barriers to innovation
Before launching the accelerator, the city government commissioned Demos Helsinki, the think tank, to undertake some research among staff. This identified specific barriers to adoption of experimentation as an approach, including lack of time, resources, and experimentation skills. The city therefore provided information and training about AI for anyone who was interested. Support and help was also available in ‘experimentation clinics’, to help employees think through ideas. This ensured that the barriers to submitting ideas for good potential projects were low. Each of the seven projects chosen for further development was given 10,000 euros from the city’s digitalisation budget as development funding.
5. There are some important fundamentals for AI projects, including transparency and user-friendliness
One of the big issues with AI is the ‘black box’ nature of some algorithms. The city views transparency as a very important principle. It recognises that acceptance of AI requires people to understand how algorithms are making decisions. It is also essential to ensure that the algorithm continues to operate appropriately. Similarly, user-friendliness is considered a key factor, because nobody will use a service that is difficult to operate or illogical. This is partly why all the projects involve co-creation with users as well as technology experts.
6. The city has involved business partners in planning and implementation of the projects
The employees generating ideas for the Experimentation Accelerator are not AI experts. The city has therefore partnered with several companies to support the development and delivery of the experimental projects. In the planning phase, partners included Deloitte, Digitalist Group, Aiwo.ai, Microsoft and SAS. Partners involved during implementation included SAS, Deloitte, Integrity and Helsinki Intelligence. The partners have been involved in both developing and assessing the projects.
