The City of Helsinki has recently developed internal innovations with what it calls an Experimentation Accelerator. The accelerator project was launched in the autumn of 2019 with a campaign that encouraged city employees to brainstorm and propose agile experiments that make use of artificial intelligence. During the winter, the city completed seven experimental projects that explored AI’s potential to improve the city’s services. Diverse experiments were made, for example, in AI-supported analysis of data from the city’s feedback system and improved targeting of fire inspections.
The premise for the City of Helsinki’s digitalisation development is that services can be digitalised to make the everyday life of Helsinki residents more fluent and to help them find services. The same applies to city employees: digitalisation helps their daily work through automation of routine tasks. In this context, AI-supported service recommendation bots, feedback analysis assistants and targeted fire inspections could hold a key role in the future.
“By means of the Experimentation Accelerator, we strive to refine the innovations and support the agile experiment activities that make use of digitalisation within the city organisation. The objective is also to learn from the experiments and scale them in order to make them available to the entire city”, says Programme Manager Ville Meloni, programme manager at the City of Helsinki’s digitalisation office.
AI supports identification of pedestrian crossings and recommendation of cultural services
A couple of examples of how machine learning can be utilised are the identification of pedestrian crossings and the recommendations bot for cultural services. Through the automatic identification of pedestrian crossings, it will be possible to support, for example, the development of smart traffic in Helsinki. The smart recommendations bot for cultural services called Löytö (“Discovery”) will help bring cultural events closer to residents by suggesting services that they might enjoy, based on their interests.
GIS Specialist Fanny Taxell headed a project to identify all the city’s pedestrian crossings from aerial photos and get their coordinates.
“We learned that the quality and volume of the teaching data is very important. The risk of failed identification decreases with growing data volumes. Teaching a model requires a lot of marked data, or in our case, images of pedestrian crossings – and for teaching purposes images with no pedestrian crossings, too,” says Taxell.
Sari Lehikoinen, who has been designing the recommendations bot for cultural services, says that the city’s services have many user groups, which is why guiding them to the services is often difficult in digital channels. On the other hand, if the recommendations algorithm can be successfully applied to cultural services, then the recommendations function can be applied using the same technology to the marketing of services in other divisions as well.
“By means of AI, we can bring the services closer to the customer, improve the findability of the services and shorten the customer journey. The findability increases the satisfaction with the services as well. At the same time, we can think about how we could offer nice surprises and the joy of discovering new things to residents,” says Lehikoinen.
The transparency in the city operations must be paid attention to, especially in the recommendations services: the residents may become interested in why a certain event and service is recommended to them. The algorithm must be transparent as well.
The creators of the AI experiments have noticed that the artificial intelligence is precisely as intelligent as its teacher. A human being is needed for teaching the machine and the biases in the teaching data or the teachers thinking are also visible in the action of the AI. According to Lehikoinen and Taxell, it is also important to look at the AI-based product or service from different perspectives right at the beginning of the project. Out of these perspectives, the customer perspective is the most essential, in order to make the result a clear and user-friendly product. In the end, no one is going to use a difficult and illogical service.
The operations of the Experimentation Accelerator continue with new AI experiments this spring
Several companies in the field have participated in the Experimentation Accelerator operations as partners to the city. In the planning phase, Aiwo.ai, Ai4Value, CGI, Deloitte, Digitalist Group, Futurice, Gofore, Helsinki Intelligence, Microsoft, SAS and Taito United offered support. The partners in the implementation of the experiments were Deloitte, Helsinki Intelligence, Integrify and SAS.
The city’s second AI experiment and software robotics campaign starts at the Experimentation Accelerator in May 2020. The experiments have been financed from the budget of the city of Helsinki’s digitalisation programme.
Further information on the city’s digitalisation programme:
click –> here