When addressing the issue of mobility in cities, it is vital to gain a firm understanding of current conditions to inform the development of technology, infrastructure, or public policy solutions. Modern cities have thrived because of mobility innovations; we would all still be commuting on foot or riding horses if it weren’t for the railway and the car (Wegener 2013). While these mobility modes have improved quality of life and access to resources both within and outside of cities, they have also allowed people, retail outlets, and service providers to spread across great distances.
These notes outline current mobility modes and their challenges, and the importance of density and proximity in urban communities. Further discussion will illustrate how fine-grained data related to complex human behavior can inform the design of mobility systems, and how technological innovation can dramatically improve mobility patterns in cities.
2. Mobility revolutions
Before we explore new mobility solutions, it is important to have an overview of the history of urban mobility. For thousands of years, transportation was largely limited to human, animal, or wind power. This changed dramatically in the late 19th century, when industrialization attracted millions of rural people and immigrants to urban areas, and triggered a period of extreme urbanization. The city of Boston, Massachusetts, for example, more than doubled its population between 1880 and 1920 (Global Boston n.d.). To accommodate this unprecedented urban growth, cities turned to the emerging technology of that time.
2.1 The first mobility revolution
In the 19th century, steam-powered, horse-drawn, and cable-hauled trams had been deployed in cities in a limited way, but the electric-powered tram was the first urban mobility technology that proved to be rapidly scalable. In 1880, the first electric tramway was built by Fyodor Pirotsky, a Russian inventor. In 1881, a commercial electric tramline opened in Germany (Train History n.d.). Soon, electric trams and subways were operational in cities across Europe and the U.S., providing highly efficient mobility to expanding cities. In New York City, elevated electric trains were quickly deployed in the 1880s and soon covered much of Manhattan and Brooklyn. Cities expanded in linear directions along these mass transit routes, while largely maintaining the high-density neighborhood patterns found in older sections of the city.
2.2 The second mobility revolution
Although automobiles had existed since the 1880s, they were mostly expensive luxuries for the affluent. This changed in 1908, when Ford introduced the Model T. This mass-produced car proved to be reliable, easily maintained, and relatively low cost. After the end of World War I in 1918, the national expansion of roadways and affordable cars for the middle class led to a tremendous growth in automobile ownership. The number of registered drivers almost tripled to 23 million during the 1920s (U.S. History n.d.). By the 1950s, electric trams were largely replaced by gasoline-powered mobility modes, enabling the low-density, auto-centric expansion of the suburbs. Urban streets became pathways for automobiles, rather than places of human interaction.
While the automobile provided many millions of people with the freedom to live and work where they wished, it also led to traffic congestion, sprawl, and serious environmental impacts, including greenhouse gas emissions and toxic air pollution. Coinciding with the spread of modern (Euclidean) zoning regulations ¬– which created separated, single-use zoning districts — the widespread use of combustion-powered vehicles has resulted in long commutes and increased social isolation. This has had negative impacts on health, wellness, and quality of life. In the 1970s and ’80s, the second mobility revolution accelerated the fragmentation of communities and the decline of many of the great urban areas. In recent years, it has become clearer that this — our current mode of urban mobility — is unsustainable.
2.3 The third mobility revolution
We are now entering a new mobility era in which three innovations may converge: autonomous, shared-use, and ultra-light electric-drive vehicles. Autonomous vehicles offer the possibility of dramatic improvements in safety, more efficient flow of traffic, and greater access to mobility services (Lang et al. 2017). Shared-use systems can eliminate the need for parking in central cities and provide people with door-to-door mobility, without the need to own a personal automobile. Ultra-light electric-drive vehicles — with less need for the heavy protection required for human-controlled automobiles — can respond to the fact that most trips in a city are by one person travelling a short distance, at a relatively slow speed. This combination will, over time, enable space now used by automobiles to be reclaimed for human activity and more productive functions. If these technologies are properly deployed, and if zoning is modernized, this new era will enable higher density development without the usual traffic and parking problems, helping to address shortages in housing availability and improving urban vibrancy. Lanes devoted to mobility can be reduced, thereby expanding space for cafés, greenery, and other high-value uses. Many crosswalks and traffic signals can be eliminated, creating shared space for pedestrians, cyclists, and a wide range of autonomous mobility modes.
The third mobility revolution coincides with a return to the city by young professionals, retirees, artists, and corporations who value vibrant urban centers. As we will see, in order to create sustainable change, technological innovation must be based on an understanding of history, human behavior, and human values.
3. Current mobility modes
The use of mobility modes varies greatly from city to city. In Los Angeles, California, the majority of people travel exclusively by car. In New York, most people walk or take the subway. In Melbourne, Australia, many people take the tram to work. In Amsterdam, Netherlands, a significant portion of urban travel occurs by bicycle. In Johannesburg, South Africa, many people use minibus taxis to commute. Although individual preferences vary, some mobility modes are preferable to others because of the external impacts on public health, the environment, and congestion in the network. For example, replacing car trips with cycling benefits the individual travelers by reducing their risk of cardiovascular disease, cancer, and all-cause mortality (Doorley, Pakrashi, and Ghosh 2015). At the same time, the displacement of car trips benefits society because of the reduction of greenhouse gas emissions and pollutants.
3.1 The challenges of current mobility modes
This section dissects the challenges that come with different mobility modes and patterns in modern cities, as well as solutions that are currently being implemented.
Although cars are the preferred mode of transport for most commuters in the developed world, they are also the biggest contributor to many of the problems faced by cities, including congestion, air pollution, noise pollution, and depletion of fossil fuels.
Congestion is one of the most common transport problems in modern cities. Since the end of the 20th century, the amount of time commuters spend in their cars has increased significantly (Rodrigue, Comtois, and Slack 2017). This is largely due to the continued expansion of cities and their inability to keep up with the demand for road infrastructure.
Figure 1: The average time and distance traveled from home to work in the U.S. from 1977 to 2009.
(Adapted from: U.S. Department of Transportation, BTS)
Why do you think the number of cars being produced and driven has greatly increased over the years? Is it the convenience of the mode? Is it because people live further away from work? Could it be because of the societal norms and patterns that accompany this invention?
André Gorz’s 1973 essay on how cars have taken over cities seems even more relevant today than it did when it was written. As you read it, consider some of the questions above. Feel free to discuss your thoughts with your fellow participants, by navigating to the class-wide discussion forum on the Online Campus.
Car traffic is a major contributor to harmful levels of pollution in cities. In London, England, alone, air pollution is estimated to kill 9,500 people each year (Vaughan 2015). In India, heavy traffic congestion combined with old vehicle technology has resulted in levels of pollution considerably higher than London’s. While road traffic is just one contributing factor to air pollution (along with factories, smoke from cooking and heating, crop burning, and dust from construction sites), it is a major contributor in many cities (Musaddique 2018). According to Stanway (2017), more than 80% of the carbon monoxide in China’s air is due to cars.
Air pollution is not the only harmful environmental stressor caused by road traffic. Noise pollution is something that often goes unaddressed when it comes to examining mobility challenges in cities. According to the European Environment Agency (2017), road traffic is responsible for the majority of noise pollution in Europe and, after air pollution, is the most damaging environmental stressor. Loud noises (above 55 decibels) trigger a stress response in the body which, over time, can lead to sleep deprivation, heart disease, and hypertension (European Environment Agency 2017).
Even though hybrid and fully electric cars are becoming more widely used, most cars are still powered by gasoline or diesel, which are unsustainable fuel sources (Shankleman 2017). The supply of fossil fuels on Earth is limited, and burning these fuels is bad for our health and the environment. If we don’t develop sustainable alternatives before the supply runs out, our energy needs will not be met, leading to serious consequences for economies and the safety of people around the world (Dayalu 2012).
3.1.2 Mass transit
Using public transportation is, of course, a more sustainable alternative to driving a private car. Effective public-transport networks can alleviate congestion in urban centers, and can be more cost effective for commuters. They provide access to employment opportunities and services to lower-income groups who may not own cars, and, if widely adopted, reduce air pollution in cities. This mode does, however, also come with its unique set of challenges for urban planners, local governments, and commuters.
Mass-transit infrastructure takes many years to plan for, and is expensive to build, operate, and maintain. Mass-transit vehicles are subject to delays, and can be overcrowded during peak hours (Narayanaswami 2016). As low-density urban sprawl leads to increasingly decentralized cities, public-transport infrastructure typically does not respond to the increased distances between the residential and commercial areas of a city, and can fail to serve all areas of an urban region (Rodrigue, Comtois, and Slack 2017). This can also result in trains and buses being underused in suburban areas, which makes these systems financially unsustainable. Future cities must develop new strategies to address these issues and provide effective and affordable mobility to all its residents.
3.1.3 Non-motorized transport
Although walking and cycling are essentially free and have excellent public-health benefits, the percentage of people who walk and cycle remains low in many modern cities. One of the main deterrents to urban cycling is the perception that it is less safe than other modes of transport. Urban cyclists are generally at a higher risk of being injured in traffic collisions than motorists are. Many motorists do not yield for pedestrians or cyclists and, often, safe pathways for non-motorized transport are not provided in the urban design (Rodrigue, Comtois, and Slack 2017). However, research has shown that there is a “safety-in-numbers effect” associated with walking and cycling, whereby the collision risk of any pedestrian or cyclist decreases as the total number of pedestrians or cyclists in the city increases. This effect has been attributed to the reduction of cars in the network and the increased awareness of motorists to the presence of non-motorized modes (Elvik 2009).
Walking and cycling are fast, efficient, environmentally-sustainable, and healthy ways to move around a city. Urban planners can encourage people to use these modes by building better sidewalks and bicycle lanes, and by creating car-free zones in urban centers or shopping districts.
Pause and reflect:
Do you think that adding bicycle lanes and associated infrastructure actually increases the number of commuter cyclists in a city? What other interventions or considerations need to be factored in? Consider these questions as you read this article about how bicycle lanes have had little success in Johannesburg.
Europe is ahead of the United States in this regard. In the U.S., longer distances between homes and amenities, cheaper cars, and the nature of public policy and infrastructure systems often make walking or cycling inconvenient and unsafe. In Europe, cities tend to be more compact and cycling is a way of life for many people, especially those living in the Netherlands or Sweden (Pucher and Dijkstra 2003). Additionally, cycling in European cities is actively encouraged through a wide range of cyclist-friendly policies, such as the provision of segregated bicycle lanes on busy roads, traffic calming in residential areas, restrictions on vehicle use in certain urban areas, and rigorous traffic education (Pucher and Dijkstra 2003).
In U.S. cities, improved infrastructure for pedestrians and cyclists could significantly improve public health, community interaction, and freedom of movement in urban areas.
Figure 2: The percentage of people who walk and cycle in North America and in Europe (1995).
(Adapted from: Source: Pucher and Dijkstra, 2003)
3.1.4 Public spaces
Roads, railways, and other mobility infrastructure can have both positive and negative effects on public spaces within a city. Parking infrastructure, in particular, typically compromises human interaction and the quality of public space. In central Houston, Texas for example, surface parking and parking garages occupy 25% of the land area, and the total area devoted to automobile infrastructure (streets plus parking) accounts for almost 65% of the land area (Gardner 2011). In addition, cars are stationary for 95% of the time on average, making parking an extremely wasteful use of valuable urban infrastructure (Barter 2013). In most U.S. cities, a significant percentage of traffic congestion is caused by motorists “cruising” at low speed while looking for parking spaces (Rodrigue, Comtois, and Slack 2017). On-street parking also consumes valuable road space, restricting the area on busy streets that could be used for automobiles, bike lanes, green spaces, sidewalks, cafés, or other high-value uses.
Public spaces in the early 20th century accommodated recreational activities, markets, parades, and other community interactions (Rodrigue, Comtois, and Slack 2017). The transition from human-oriented public spaces to streets designed for machines occurred gradually over the 20th century. Video 1 shows a scene in Barcelona, Spain, in 1908, filmed from a newly deployed tram. Watch the first minute or two, and notice how the street space is shared by trams, pedestrians, cyclists, carriages, and automobiles. The city at that time did not have dedicated car lanes, bike paths, sidewalks, stop signs, traffic signals, or crosswalks. Humans and machines coexisted in a shared public realm in a manner still found in many older European cities today.
Video 1: Barcelona en tranvia (Barcelona tram).
We are now at the dawn of the third mobility revolution, transitioning from human-controlled and privately-owned vehicles to a wide range of new mobility modes and service models. The next section provides an overview of some of the emerging mobility solutions currently active in cities around the world.
3.2 Current mobility solutions
For many decades, transportation engineering was singularly focused on increasing the car-carrying capacity of urban and regional networks, by building new roads or adding lanes to match increased demand. This addressed the problem in the short term, but cities found that additional capacity was quickly matched by additional use. Typically, congestion returned to previous levels within about five years of major improvements, but with much higher numbers of cars using the route, as suburbs expanded and commuting distances increased (Duranton and Turner 2011). This phenomenon is now considered a fundamental law of road congestion. However, new mobility options are emerging that may alter the auto-centric urban development typical of cities around the world. Many of these solutions are based on sharing, rather than private ownership.
Car-sharing is a type of car-rental service that allows people to rent vehicles by the hour, or for other short periods of time. The service is used primarily by customers who make only occasional use of a vehicle, or who require a different vehicle for special purposes. Founded in 2001, Zipcar, a subsidiary of Avis Budget Group, is an American car-sharing company that allows users to reserve a car through a smartphone app and unlock the car with the card they are provided with when they subscribe to the service. They can use the car for anything from one hour to two days, and then park it at an approved drop-off location. Conveniences such as this might one day deter people from purchasing a car they will only use on weekends or once or twice during the week. Local governments incentivize the use of this service by providing designated parking bays for car-sharing only. However, because car-sharing and ride-sharing services make it cheaper and easier than ever for people to travel by car in a city, some studies have suggested that these services are leading to more congestion in cities such as San Francisco (Marshall 2018). The growth of car-sharing is constrained by the even more impressive success of ride sharing services over the past decade.
Ride-sharing companies provide taxi-like services by connecting passengers to drivers who use their own vehicles. Uber, founded in 2009, and competitors such as Lyft and DiDi (China) are rapidly disrupting the taxi industry, impacting the automotive sector, and changing how urban residents access mobility. Many consider ride-sharing to be a viable and more convenient alternative to private car ownership for urban mobility, and believe it may play an important role in the reduction of parking required for central cities. But recent studies have also shown that ride-sharing companies are adding car trips to urban streets and, in many cases, stealing passengers from subways and buses. One study showed that more than 50% of ride-sharing trips would have been made by transit, walking, or bicycle — or would not have been made at all — if ride-sharing was not available (Clewlow and Gouri 2017). The major ride-sharing companies are anticipating a future where drivers using their personal vehicles are replaced by fleets of autonomous robo-taxis.
Public and private bike-sharing systems have been successfully implemented in many cities around the world. The largest bike-sharing programs include Bixi Bikes in Montreal, Canada; Vélib’ in Paris, France; Forever Bikes in Shanghai, China; Santander Cycles in London; and Citi Bike in New York City. Beijing-based Mobike, however, has taken bike-sharing to the next level, by using Internet of Things (IoT) technology to develop station-less bikes. Riders use an app to unlock the bicycle and then park it anywhere once they are done using it. This avoids having to find a bike station at the beginning and end of your journey. Bike sharing is a quick and convenient way to get around a city, without having to own a personal bicycle or worry about theft. Bicycling promotes public health through increased physical activity, and can enhance community interaction. In many cities, the growing number of bike-share users has resulted in effective advocacy for safe bike-lane infrastructure.
3.2.4 Bus rapid transit (BRT)
BRT is a mobility solution that is experienced much like an underground metro system. However, these buses operate on existing street networks, thereby negating the time, expense, and disruption required for the installation of tunnels and underground stations for a subway system. BRT differs from a standard bus system in that buses drive in segregated lanes, stop less frequently, and are synced to traffic signals so they spend less time waiting at intersections. BRT, when properly designed, creates a faster and more seamless commuting experience for passengers. BRT stations also function more like metro stations, as riders enter the station through a fare-controlled zone and access the bus with the platform and bus floor at the same level. Many have sophisticated ticketing systems (Elledge 2016). Because BRT buses tend to be high capacity, they can move the same number of people with fewer vehicles than traditional bus systems, reducing traffic congestion and air pollution (King 2013). Although BRT systems have significant cost and operational advantages, their implementation has been met with controversy in cities such as Cape Town, South Africa, where local taxi drivers have protested against their loss of revenue, and car owners have objected to dedicated BRT lanes in favor of more lanes for automobiles instead.
We can see from these examples that promising new mobility services may have unintended consequences, with both positive and negative impacts on a city. An alternative way to address the problems of traffic congestion and parking is by reducing the need for commuting in the first place. As we will see in Module 3, the demand for travel may be dramatically reduced by creating higher-density, diverse urban districts where people can live, work, play, and socialize within a compact, walkable neighborhood. In addition, new mass-transit and personal, autonomous mobility solutions may enable significantly more efficient use of public infrastructure.
New mobility modes and urban infrastructure will ultimately influence the individual decisions made by thousands or millions of people. Influencing choices in this way requires a nuanced understanding of mobility behaviors and motivations. For this reason, we also need to build predictive models of mobility behavior that are accessible by planners and community members alike. In the casebook for the next unit, you will explore some of these emerging mobility solutions that could have a lasting impact on how people move around cities, if implemented successfully.
4. Understanding mobility patterns with data
4.1 Dimensions of mobility behavior
Many forms of data can be used to help us understand mobility patterns through analysis or building of statistical models. Transportation data sources have traditionally been analyzed at a systemic level, and most cities still use this approach. Cities are divided into zones and transportation behavior is analyzed at an aggregate level using the traditional four-stage model of generation, distribution, mode choice, and traffic assignment (Ortúzar and Willumsen 2011). This is known as the aggregate approach.
Figure 3: The classic four-stage model.
(Source: Ortúzar and Willumsen 2011)
The disaggregate approach, on the other hand, recognizes that the mobility patterns of a city are the result of many decisions made by individuals. The disaggregate approach has several important advantages over aggregate methods. While aggregate approaches are primarily statistical in nature, disaggregate approaches directly explain why an individual makes a choice given their circumstances. Those analyzing the data are therefore better able to predict how those choices may change in different circumstances (Koppelman and Bhat 2006). Additionally, because disaggregate models represent the behavior of each individual, they allow us to visualize movement patterns in the city more effectively and to make more precise estimates of, for example, exposure to pollution as travelers make their way around the city. As discussed later in these notes, the mobility choices made by each individual can be further broken down into decisions regarding activities and locations, mode choices, and route choices.
We often talk about transportation in terms of supply and demand. The supply side requires a network infrastructure that is capable of meeting the demand for transportation. Transportation provides a means to participate in activities in different locations, so the demand for transportation is derived from the demand for activities. The first step in understanding mobility is therefore to understand how and why people choose which activities to participate in and where to do so. The series of activities performed by a person throughout a full day (for example, traveling from home to work and back home again, or from home to work and then to the gym, before returning home again) is often referred to as a “daily activity schedule” or “mobility motif”.
Given a series of activities and locations, people must then choose which mode of transportation to use for each trip. As previously discussed, the most popular options include driving alone, carpooling, ride-sharing, cycling, and walking. In the future, this list of options may expand significantly. The choice of mode usually depends strongly on the length of the trip, the demographic group of the commuter, and the trip purpose. Finally, for each trip, the traveler must choose which route to take through the transportation network. Route choice is relatively well understood for drivers who tend to follow the route with the shortest time, with trade-offs for tolls and other monetary costs. However, for non motorized modes, route choices can be less predictable, as pedestrians and cyclists often choose to take routes that are more scenic, less risky, or less polluted.
4.2 Data analysis and model development
In many areas of urban science, we use models to better understand complex systems or predict how they may respond to changing conditions or interventions. These models are simplified mathematical representations of real phenomena and, in order to be useful, their behavior should closely approximate the behavior of the real system. In Unit 3 of this module, you will learn about agent-based modeling (ABM) and how this can be used to understand, visualize, and predict urban mobility behavior. In ABM, simulated agents have to make decisions, such as which activities to participate in, where to pursue those activities, and how to move between them. Often, agents will have different characteristics or profiles that will influence the choices they make. If no suitable data are available, these decision-making processes may be codified subjectively by a modeler with substantive knowledge. However, if data are available, the agents’ profiles and behaviors should be calibrated using it in order to ensure they are realistic.
In this section, we will outline a simple example of how to use activity-based analysis (ABA) to calibrate such a model based on data.
4.2.1 Data requirements
The first step in calibrating a disaggregate model is to gather data about the choices made by individuals and the variables that we expect may have affected those choices (Koppelman and Bhat 2006). Typically, this would be in a spreadsheet format where each row represents one person and each column represents a variable for that person. The following are some of the variables that may be used:
- Traveler and trip-related variables such as income, car-ownership, etc.
- Alternative specific variables such as the travel time and cost for each mode
- The actual choice made by the traveler, for example, driving
(Koppelman and Bhat 2006)
One example of a suitable data set is the National Household Travel Survey (NHTS) 2017, an open data set about travel behaviors across the U.S. The NHTS data contains hundreds of variables about each person surveyed and each trip they made on survey day.
4.2.2 Model development
In this example, we will model two types of choices, the mobility motif, and the mode choice, both of which are captured in the NHTS data. There are numerous model structures we could use for each of these choices, the most popular being econometric or rule-based structure. For the sake of this example, we will employ a rule-based approach known as a decision tree, which is relatively simple and has been used successfully in many contexts. A decision tree is a type of machine-learning model that can take observations about an item and use them to simulate a decision-making process that leads to an outcome. The decision tree must be trained (calibrated) using data so that it accurately reflects the choice-making behavior of the population.
Figures 4 and 5 show decision trees for mobility motif and mode choice that have been trained using the NHTS data. In order to maintain simplicity, the depth of the trees has been limited to three options, so that only a small number of the most important characteristics are used to simulate the decision-making. Figure 4 shows that the model first checks to see if the agent is a homemaker. If so, the agent will be assigned to Profile 1 and will have a 66% probability of having the “H” motif, that is, a stay-at-home parent. If the agent is not a homemaker, other attributes (age, vehicles available, and education level) are taken into consideration in deciding on a profile. In Figure 5, we see that trip distance has turned out to be the dominant factor for determining the mode of transportation. Trips of less than 0.28 miles are likely to be walked or cycled; but, for trips greater than 0.28 miles, other attributes such as whether there are school-aged children in the household are also considered in determining the mobility mode. It should be noted that, if a different data set were used, such as data from a different country, the decision trees would likely turn out differently.
This example is highly simplified. In practice, larger trees with more variables may be used and techniques such as pruning, cross-validation, and boosting would need to be applied in order to ensure robustness of the models, but such techniques are outside the scope of this program.
Figure 4: Decision tree for determining profile and motif probabilities for simulated agents.
The letters “H”, “W”, and “O” refer to sequences of “Home”, “Work”, and “Other” activities over the course of a day.
Figure 5: Decision tree for determining mode of transport for simulated trips by individuals.
4.2.3 Cell-phone data
The individual-level disaggregated survey data described in the previous section can be more difficult and expensive to obtain than aggregated data, which is one of the main reasons why aggregate models are still popular in many cities. However, recent research has been exploring how cell-phone data can be used for disaggregating mobility models in place of survey data. Cell-phone penetration is close to 100% in most developed countries, and well over 90% in many developing countries. The locations of cell phones are continuously monitored by telecom providers, so they offer a rich source of passively collected data for disaggregate mobility analysis.
Call detail records (CDRs) are records kept by telecom service providers containing details of each time a subscriber used their phone for a call, text, or data; and which cell tower they were connected to. Through knowledge of the approximate coverage regions of each cell tower, we can reconstruct the approximate trajectories of every subscriber. The CDRs may also contain other information about the subscriber, such as their home network (if roaming) and their device model. The coverage region of each tower is often large or difficult to approximate, which means that the spatial resolution of CDR data can be coarse compared to, for example, GPS records. However, it is possible to combine observations about a single device from multiple towers in order to “triangulate” the coordinates of the device. Many telecom companies do not retain this level of data, so CDRs are more commonly used for mobility analysis. Some examples of mobility models based on CDRs are provided.
The City Science research group at MIT, developed an agent-based model representing mobility patterns of tourists in the micro-state of Andorra based on CDRs. This ABM allows Andorran stakeholders to observe the behavior patterns of tourists from different countries on normal days, as well as on the days of major events, such as the Tour de France or Cirque du Soleil. You will see a demonstration of this model in Unit 3.
Jiang, Ferreira and González (2017) used CDRs in Singapore to recognize and extract individual mobility motifs, similar to those discussed in section 4.2.2. They also validated their patterns by comparison with household travel survey data. Approaches such as these could eventually reduce or eliminate the need to conduct expensive household travel surveys that quickly become outdated.
Like many other forms of “big data,” the analysis of cell-phone data brings up obvious privacy issues. Even if the data are anonymized and any identifying information removed, simply having knowledge of a user’s location history could potentially be used to identify them because of inferences, for example, to their home location. Various techniques may be used to transform the data to ensure that details such as home or work location cannot be inferred, but these usually entail some loss of resolution or accuracy in the data.
Rather than obscuring or aggregating the data before they are released to the modelers, an alternative is to bring the model to the data. The overview of open algorithms (OPAL) architecture (Pentland et al. 2016) describes a framework for training models using sensitive data without the raw data having to leave their home repository. In this framework, the training algorithm must be made open to the public where it can be vetted by experts to be safe from discrimination or privacy violations. The vetted algorithm can then be applied to the data in their home repository and the results can be returned to the author of the model or algorithm. This framework allows the modeler to achieve their goals — training a model with individual data — without ever gaining access to the sensitive data.
In these notes, we discussed current mobility patterns around the world and some recent attempts to solve them. Although there have been many technological and service-model innovations in recent years, it remains to be seen whether these will positively impact cities in the long term. In order to ensure that innovations lead to sustainable improvements, they must be coupled with an understanding of human behavior in relation to mobility choices. Such an understanding helps us to design appropriate interventions and to predict how people’s behavior will change with the introduction of a new mobility system. In Unit 2 of this module, we will discuss ongoing work in developing next-generation solutions to urban mobility, such as lightweight autonomous vehicles and new kinds of mass transit.
We have also seen a brief overview of methods for understanding and modeling mobility behavior at an individual level using data. Both traditional forms of data from surveys and emerging forms of data, such as cell phone-use activity, can be useful in understanding mobility behavior. Unit 3 will focus on how we can use simulation to predict how our understanding of current conditions and behavioral patterns is likely to change in response to the introduction of new mobility systems or other interventions.