Using MPC to generate relevant insights to support inclusive travel

Can you encourage people who depend on WMO (Social Support Act) transport to choose public transportation more often if you offer them a free public transport pass? This was the central question of the MaaS (Mobility as a Service) pilot ‘Inclusive Travel’, which was set up by our participants Publiek Vervoer and Roseman Labs. We spoke with Petra Buitenhuis, Project Manager at Publiek Vervoer, and Freya de Mink, Business Developer at Roseman Labs, about the societal relevance of this pilot, the challenges of data sharing, and the relevance of Multi-Party Computation (MPC) as a solution to address these challenges.

 

Combining travel data from different data points

Publiek Vervoer is a collaborative organisation of twenty-two municipalities in the provinces of Groningen and Drenthe. It focuses on target group transportation such as WMO transport, student transport, and transportation to social workplaces. In collaboration with the Public Transport Authority, it ensures a comprehensive and well-functioning transport network in the region. Petra: “We believe it’s important to integrate public transport and target group transportation as much as possible and to ensure accessibility for all target groups in our national provinces. In various pilots, we are exploring how we can implement Mobility as a Service (MaaS) in the region.”One of these pilots is ‘Inclusive Travel’, which examines whether people in seven municipalities who normally use WMO taxis are open to choosing public transport instead. Petra explains: “There are several reasons why this is important. Social inclusion is a top priority. We believe it’s crucial that people can participate in regular life as much as possible, which includes traveling by public transport. Additionally, WMO transport is becoming increasingly expensive. Costs are projected to rise by up to 30% if no other policy choices are made. Encouraging and assisting people to choose regular public transport is a solution to limit costs. Moreover, it can be an argument to at least maintain the same frequency of bus transport, which is already under pressure, and ensure accessibility throughout the province. We are pleased that over 300 people participated in the pilot.”Only people that are entitled to the WMO and eligible to use WMO taxis participated in the pilot. They received a special public transport pass that allowed them to travel for free on public transport during the pilot. The hope and expectation was that they would rely less on WMO taxis at the end of the pilot compared to the beginning. However, according to Petra, this hypothesis was not easy to investigate for several reasons. “Travel data is considered personal data under the GDPR. Institutions are not allowed to share such data without permission. Additionally, crucial data for the research is stored in multiple locations. This includes travel history data at the taxi company and demographic data such as place of residence and age across the seven different municipalities. Furthermore, not every municipality had structured the data in the same way. Therefore, the major challenge was how to combine the useful information from all these data points.”

No data standardisation required in advance of a MPC project

Publiek Vervoer connected with Roseman Labs, who provided assistance to map the results using their MPC technology. Freya: “With our technology, we help organizations collaborate on sensitive data. This starts with encrypting the data at the source. MPC then involves distributing encrypted data fragments to three different servers hosted by European cloud providers. These servers, known as ‘parties’, together form the privacy engine where calculations take place.” The eight data points (the seven municipalities and the taxi company) were able to do this easily using a module of Roseman Labs’ software. Freya: “We perform several quality checks beforehand, such as checking the file size. If it’s too large, the municipality officer receives an error message.”However, the data analyst doesn’t have to depend on the data source itself for every adjustment. “The biggest misconception about MPC is that all data sources must adhere to the same structure and standards before a project can start. While it helps if parties adhere to the same standards as much as possible, it is not a requirement for MPC. For example, consider the data related to passengers’ dates of birth. Some municipalities use the structure ‘year-month-day’, while others use ‘day-month-year’, and a few use separate numbers. If an analyst performs a calculation on that data, it would obviously be incorrect.That’s why we included a quality check that reveals one field from each source while the surrounding data remains hidden. This allowed us to determine if the structure was correct or not. With that knowledge, we wrote code to assemble the correct sequence, ensuring the data was structured in the same way. We can provide very specific instructions because we work with a Python-based interface. The flexibility it provides saves a lot of time as we don’t have to go back to the source to make changes to all the data.”Petra confirms the added value of MPC: “We often conduct research where municipalities need to provide us with data lists, and they all use different back-office systems with different formats. The flexibility offered by MPC reduces our dependency on the structure chosen by the municipalities and helps shorten the research time.”

Complying with GDPR guidelines by using MPC

According to Freya and Petra, MPC technology is also a good way to mitigate risks in line with GDPR requirements. Freya: “The major advantage of MPC is that data is encrypted at the source and remains encrypted at all times during processing. Only results are revealed. You might be inclined to think, ‘give me as much data as possible because we might be able to use it later on in the research.’ However, a key requirement of the GDPR is to minimise the data you process and only request the data necessary to test a hypothesis. For this research, we requested people’s dates of birth but not their names. ‘Purpose limitation’ is an important principle: you can only use the data for what you have agreed upon with the data owners.”Petra adds: “Together with the participating municipalities, we conducted a Data Protection Impact Assessment upfront to identify and mitigate potential risks ” The technology ensured that the research outcomes couldn’t be traced back to individuals and that passengers’ privacy would not be compromised. “Although travel history at the individual level was used to determine changes in behaviour for specific groups, the analyst never had access to this individual level data.” Petra explains. “For instance, we always worked with buckets of ten people belonging to the same age group. The aim of the research was to determine the average impact of the public transport pass on our target group, not its individual-level effects.”

Why more travel data is needed to verify the results

MPC technology has proven its value, but did the experiment with the free public transport pass also yield the expected results? Petra: “We have to be cautious, but we can be cautiously optimistic. The pilot started in 2019, just before the COVID-19 pandemic. Naturally, due to the subsequent measures, we saw a dip in taxi transport. We can accurately identify when these measures had the greatest impact on travel behaviour. Interestingly, there is a significant decrease in the use of WMO taxis at the end of the pilot compared to the start in 2019. This is particularly evident in urban areas where public transport facilities are more accessible.” However, it remains to be seen whether more people started using public transport during the same period. Petra: “Our hope is to add train and bus data to this pilot to measure that effect. If our cautious assumptions can be supported by this additional data analysis, it is an important signal for other municipalities to start a similar pilot.”