How big data and AI can help policymakers evaluate mobility projects
Following the vehement protests against the Good Move plan, the much-discussed Brussels mobility plan that was implemented in several phases starting in 2019 and shortly thereafter was already discarded by a number of municipalities, scientists from the VUB Data Analytics Lab and the VUB Mobilise research group decided to find out to what extent that negative vibe matched reality. An examination of Tweets in Brussels during the period between 2019 and 2022 reveals that opinions on successive implementations of the plan were not nearly as negative as one might think. The result must be viewed with a certain nuance, since the audience on Twitter (now X, nvdr) is not a representative sample of the population. It does show that on Twitter - known for a polarization bias - opinions were not overtly negative, but rather nuanced toward the mobility plan.
The trigger for the research was the heated protests against Good Move that dominated the Brussels news starting in the fall of 2022. Siblings Floriano (VUB Data Analytics Lab) and Sara Tori (Mobilise) wanted to find out how representative those protests were of Brussels’ online active population. "We started collecting Tweets for the period between July 18, 2019 and December 31, 2022 about mobility changes in the city," says Sara Tori.
The results revealed an unexpected picture of the Brussels Twitter population’s perception of the mobility plan. "The peaks in Tweet traffic were always linked to the announcements or the coming into force of parts of the mobility plan. We noticed that Twitter is thus a platform where people came to express their opinions about changes in mobility."
From news articles, it seemed that the media was chasing negative commentators. Yet the Brussels mobility plan carries more popular approval (on Twitter) than appeared from media coverage.
The results of the study indicate that social media can also be interesting polling tools for research to ascertain some aspects of public opinion. What is important here is that it is used complementarily to traditional survey techniques such as questionnaires or information evenings. "Our research also shows that using a single AI tool like GTP4 can be useful for policy makers," Tori said. "Moreover, we show that GPT can better handle contextual information. A tweet that used negative language but actually expressed a positive opinion about Good Move (for example, deploring a rollback of the plan) was more often correctly recognized as positive by GPT than by other models."
The research is also directly relevant to policymakers, and to Brussels residents in general, who can get a more nuanced picture of how Good Move is evaluated by online active citizens. Together with media coverage, this provides a fuller picture for targeted policy making.
The research was conducted in Brussels by Floriano Tori, Sara Tori, Imre Keserü and Vincent Ginis and has now been published in Data Science for Transportation. The article is entitled Performing Sentiment Analysis Using Natural Language Models for Urban Policymaking: An analysis of Twitter Data in Brussels.
