Researchers from the Data Science and Transportation groups of the Complex Engineering Systems Institute (ISCI) used previous the Institute’s earlier versions developed a portable biomonitor that measures psychophysiological data in users of different modes of public transportation, among other areas of application.
Based on the data obtained in a pilot plan with 44 people, together with the contextual information of each trip, they developed various models of both causality and Machine Learning to validate a tool that makes it possible to measure the emotional state as a proxy for the level of user satisfaction, in an unbiased and granular way.
Their experience led to the researchers being awarded a FONDEF project at the end of 2022, in collaboration with the Directorate of Metropolitan Public Transportation (DTPM), the Road and Urban Transportation Program (SECTRA), and ISCI as an associated institution, to apply this innovative tool on routes of interest and demonstrate with georeferenced records the emotional states of people. Based on these results, concrete actions will be identified that will improve the well-being of bus drivers and their passengers.
Measuring the satisfaction of public transport users, both in the capital and in the regions, is essential for the authorities of the Ministry of Transport. The objective is to evaluate measures and propose concrete actions that will raise the quality of the service and impact people’s lives such as improving the infrastructure of a bus stop, installing segregated routes on certain specific routes, or using electric buses.
Traditionally, these measurements are made via satisfaction surveys applied to post-trip users. However, this data usually has high recall biases, since the person responds based on what they evoke from their experience and the results obtained have no ecological validity.
Biases associated with prejudices are added, mediated by social, psychological, and political variables. Finally, another difficulty is that transportation systems are highly dynamic, making information inputs valid in limited times only.
Can more objective data be obtained?
Researchers from ISCI’s Data Science and Transportation groups developed a possible solution.
During a journey, a person is subjected to multiple stimuli (luminosity, temperature, shouting, braking, interaction with other users, etc.), all of which generate different emotional states, which respond to what is happening in their nervous system. The way to quantify these states is to record the peripheral signals that the person experiences with each emotion.
Based on previous experiences in which technology was used to record psychophysiological signals from people working in other industries, the group developed an innovative bracelet-type device called a biomonitor.
Biomonitor version 3.0
The biomonitor device is minimally invasive, capable of recording pulse, electrical conductivity on the skin, surface temperature, as well as inertial signals (accelerometry and gyroscope). The device delivers a streaming and time series of high frequency data of approximately 50 measurements per second.
“There were questions regarding the optimal measurement of the utility and satisfaction perceived by public transportation users, and we have the technological development and the design of experimental methods to answer them. We set the goal of identifying the variation in emotional behavior, also known as the affective states of users, depending on different conditions or modes of transport (electric bus, traditional bus, walking, connections, subway, etc.) and recording how passengers are affected by different situations they face during their journey. We call these “attributes”. For example, traveling on a segregated route, the level of overcrowding, or traveling through places considered unsafe”,
explains Ángel Jiménez, the ISCI researcher who headed the development of the first two versions of the biomonitor.
Speaking of his own experience, ISCI researcher Ángelo Guevara, who headed the current third version of the device, considers that “the human factor is a fundamental part of the analysis of transportation systems and in various areas of the economy in general. Traditional methods only allow us to account for this effect in an aggregated and biased manner for several reasons. In economic analysis, since the 19th century, people have dreamed of having instruments that allow them to continuously measure the level of satisfaction of individuals. Our project adds a grain of sand to fulfilling that desire.”
The Pilot Plan
The first pilot plan was conducted in November 2019, with 44 students taking part, divided into groups of four people, and accompanied by a member of the team as a designated “experimenter.”
Participation in the pilot plan began with the measurement of the psychophysiological signals of the participants to measure their base emotional states before starting their journey. This measurement considered particular health conditions that may alter the data. The ISCI team worked on debugging the instrument to identify participants’ baselines, and when processing the signals, each user’s baseline was subtracted to measure the variability of their emotional states during the journey.
Each group then completed a route of approximately 2.5 hours with their biomonitor connected. The selected route included several modes of transportation:
- A walk from the Beauchef campus of the Universidad de Chile to the Parque O’Higgins Metro station.
- A journey on a specific electric bus from the stop located at the exit of the Parque O’Higgins Metro station along Avenida Grecia.
- A bus-Metro connection from the bus stop at the Grecia Metro station followed by a Metro journey towards Tobalaba.
- A walk from Avenida Tobalaba to a bus stop located at the end of Avenida Pedro de Valdivia.
- A journey on a specific conventional bus from the stop at the start of Avenida Pedro de Valdivia towards Avenida Matta.
- A walk from the stop located at the exit of the Parque O’Higgins Metro station to the Beauchef campus of the Universidad de Chile.
The experimenter had, in turn, a device called Contextino that captures the environmental conditions of the service, such as CO 2 levels, vibration, electromagnetic field, air humidity, noise level, and ambient luminosity.
All the data obtained is complemented with self-reports of the users’ emotional state during the journey. With this objective, the researchers developed an application for smart phones that randomly posed questions to the pilot participants. To answer, they had to choose their emotional state, within four quadrants, according to the Circumplex Model of Emotions.
This model classifies emotions according to their intensity and valence. In order to adjust it to the national reality, Industrial Engineering thesis student, Carlos Barría, held a focus group and a statistical analysis, which made it possible to create the adjectives for each quadrant according to how Chileans express those emotions.
Why was the self-report necessary?
To complement the information obtained from psychophysiological and contextual signals. With the complete data, the team developed various models of both causality and Machine Learning, which are capable of classifying the emotional state of public transport system users without the need to ask the person directly about their experience, visualizing the information in georeferenced maps of emotional states.
“We use the self-report as an indicator of a person’s emotion, either to estimate causal models or Machine Learning models, which relate these indicators to contextual attributes. For example, these tools allow us to measure the subjective impact that the level of overcrowding or CO 2 has on users’ perception of trip quality,”
explains Ángelo Guevara.
“To be honest, this first stage consisted in validating the instruments and methodology, but simultaneously allowed us to draw conclusions from the georeferencing of the emotional states of the pilot participants. Based on the Machine Learning-trained model, we observed that the route with the electric bus generated emotions of positive valence, represented in warm colors, which leaned towards red. While the conventional bus route generated emotions of negative valence, represented in cold colors, close to blue. We zoomed in on the connection at the Grecia Metro station and we observed a blue color, which indicates negative emotions. Why does this emotional state occur during a connection? Do users mind walking? Do they feel unsafe? We really don’t know, but the map gives you clues to go investigate on the ground exactly what it is that affects users in this way”
says Ángel Jiménez.
A new application of the concept of Instant Utility
Traditionally, the design of transportation systems has been based on classic utility models perceived by passengers, which incorporate travel time (perception of relative time) and cost as variables. In this model it is assumed that users make decisions rationally, however, this is not always the case.
Based on the results obtained in the pilot, ISCI researcher Ángelo Guevara proposed a conceptual framework that uses the concept of latent variables to model the causal relationships between granular events and emotional states. This proposal complements the classic utility model to explain the choice of transportation mode (car, bus, metro, among others). The PhD thesis in Engineering Systems, Bastián Henríquez, proposed a variation of this model that can be related to the concept of “Instant Utility” proposed by Daniel Kahneman, Nobel Prize in Economics, expert in Behavioral Economics, allowing it to be measured effectively for the first time as far as the researchers are aware.
“Instant Utility is the founding model that we have been perfecting along the way and considers the integration of utilities built based on past experiences with different modes of transportation, which the person has point by point over time. In terms of the practical application of this model for the design of public policies, obviously, knowing how a particular individual thinks is not directly usable. However, if the aggregation of micro decisions of a large number of people is analyzed, rich information is obtained regarding user perception, which allows a good diagnosis of what is happening on different routes or modes of transportation. Secondly, it enriches the models and design decisions of transportation systems, because you can identify areas that are compromised, where we see that negative valence for example, and face the problem to change that perception”
says researcher Ángelo Guevara.
Biomonitor versión 3.0
ISCI researchers Ángelo Guevara, Ángel Jiménez and Marcela Munizaga, and their team are currently participating in the new FONDEF project “Development of a psychophysiological microdata platform for the evaluation of public transport, integrating service level and perception information”, which seeks to apply the validated tool in Santiago and the Valparaiso Region.
The project contemplates that the public authority will instrument passengers on specific routes to obtain data on their instantaneous emotional states that will be complemented with passive operational and contextual data. These will be processed on a platform to draw unbiased conclusions about the quality of the system.
“The participation of the DTPM and SECTRA, representing the State, is essential for the FONDEF project to translate into a tool that can be valuable in practice. We hope that these scientific and technological developments will have a palpable impact on public transport users in the long term,”
“In the first stage, it is essential to design indicators with the public authorities to define exactly what information they want to obtain and how they want to visualize it. SECTRA, for example, needs to fully understand what is happening on the Alameda axis (one of Santiago’s busiest routes) and how they want to obtain the aggregate metrics of critical points, such as Las Rejas, General, Velázquez, Estación Central, Ahumada, and Vicuña Mackenna. Each one represents large flows of people and security issues. Our model allows us to georeference the emotional states of users at these specific points and they will be able to determine specific measures, such as modifying bus frequencies or changing the configuration of bus stops that helps to improve the experience. The DTPM, for its part, wants to apply our model to bus drivers in order to understand their psychophysiological states and to identify actions that ensure their well-being, as well as quantify how the same route affects male drivers versus female drivers. The data will be complemented with information from cameras installed on buses, and we will develop an algorithm to estimate the level of overcrowding and evasion. As a team, we believe that we have developed a truly innovative tool that will move the needle in the transportation area and implement measures that objectively impact the quality of life of users”
concludes Ángel Jiménez.