The End of the Map as We Know It
As OpenStreetMap is increasingly embraced in more commercial platforms—and quietly eschewed by some early adopters—there seems to sometimes be an atmosphere of uncertainty and even doom about the future of the world’s greatest collaborative map (by which I mean database, of course).
The End is Near - Photo by Larry D. Moore, Wikimedia Commons
As many parties race to have influence on the direction of the map, the criticisms shouldn’t be taken as an indicator of change by way of a sudden break. Things are certainly more complex now. There are over five million registered users as of April 2019, which is more than the population of countries like Ireland, New Zealand, or Costa Rica but far fewer than the number of subscribers to popular podcasts such as the Joe Rogan Experience or Serial (the latter of which hit five million listeners within months of its inception in 2014). OpenStreetMap is not a viral icon in our culture, attracting tens of millions of editors and users while trying to keep up, but with 40,000+ users editing in the last month, it spawns a significant amount of data that requires the attention of many skilled people to validate, monitor, and maintain. Is it fair to claim that the volume as well as the profile of users and their edits has not undergone any dramatic metamorphosis, but rather, an evolution?
Evolution of the Bicycle - Wikimedia Commons
Much of our technology has the same story—something small becomes larger, gets new modules, finds creative new uses, branches into different manifestations, is tugged and pulled in different directions. In the case of OpenStreetMap, its evolution has been sometimes democratic, sometimes by natural selection—tagging proposals are voted on but other times it’s a case of survival of the fittest tag. There hasn’t really been a pivot, at least that I’ve seen in my OSM use during the past four years, but instead a gradual growth that can be traced back to OpenStreetMap’s origins. It appears that OpenStreetMap’s cultivation has been an organic process, built by communities to fit their respective visions of local geography and how things on the map relate to one another.
At FOSS4G North America 2019, Paul Ramsey suggested that geospatial open source may be experiencing a midlife crisis. Meanwhile, as OpenStreetMap turns 15 years old this August, we’re seeing the growing pains of adolescence. For many, what’s at stake is whether or not OpenStreetMap’s maturation beyond this adolescent stage will remain organic. What is OpenStreetMap becoming, and why? Will sudden change actually come, and we see a living, breathing project become something of a mechanized cyborg?
There are several key pieces of the OpenStreetMap project that we need to examine. The two I’d like to call forward are the input of data into the map’s database, and the output of the database into many applications, as mentioned, from mobile navigation apps to popular websites to humanitarian projects.
The people and groups who are collecting the data have direct influence on the evolution of the project, as do the methods they use. GeoChicas, for example, is a group focused on closing the gender gap in the OpenStreetMap community, with a focus on collecting map data about centers of care for such issues ranging from to health and violence. The Humanitarian OpenStreetMap Team focuses on improving maps for humanitarian action and community development, mapping things like roads, energy grids, or access to clean water. Rideshare companies such as Grab also edit the map, improving navigation and context for pick-up and drop-off points. In other cases, individual OSM users will use satellite or street-level imagery as a reference to draw new data on the map by hand, with no automatic extraction. At Mapillary, we are extracting data from street-level imagery using computer vision, then generating point features that can be added to OpenStreetMap. Meanwhile, organizations like Development Seed, collaborating with the World Bank, use machine learning to extract map data from satellite imagery, then manually add it to the map.
Drew Bollinger recently presented Development Seed’s work on detecting electrical networks in Pakistan.
There are countless other contributors, including small groups, corporations, or individuals. Everybody deserves credit for their efforts. Everyone influences what is mapped, regardless of motivation, and shapes what the next rendering of the database looks like. Some of this data is added to OpenStreetMap because there is a demand for it, such as urgent need for mapping new roads or buildings in the wake of tropical storms in Mozambique, or commercial navigation’s demand for improved addition of turn restrictions and business locations. In other cases, such as OSM Portugal’s project to map the Azores or Bike Ottawa’s improvement of cycle infrastructure mapping, community volunteerism helps make the map better for its own sake as well as side effects like better maps for tourism or safer cycling.
Mapping the Azores with OSM Portugal.
Some of this data is collected on foot, in the field. Some is imported from government datasets. Some of it is traced from imagery, some from GPX tracks in mobile phones. An increasing amount of it is generated by algorithms that analyze telemetric data, aerial imagery, or street-level imagery. All of it is added by a human, whether out of passion or for pay, with the stroke of a keyboard and click of a mouse. Every one of us is a gatekeeper.
Aside from the who, what, and how of data input, we need to consider the same aspects of OpenStreetMap’s output. Who does it serve? In many cases, those doing the editing are the same ones driving demand.
For example, Grab is transparent about its directed editing activities, but also connects with local communities to ensure quality and collaboration. HOT both consumes the data—along with partners including the Red Cross—as well as coordinates data collection and editing. The important question to ask when considering the end-users of these maps is not what they are demanding be added to the map, but what they are leaving out. Organizations that use OpenStreetMap often tend to be interested in very specific data layers, and thus drive demand for those layers, while not addressing the layers that don’t interest them. Humanitarian organizations, for example, aren’t very interested in using street-level imagery of the Burning Man festival to improve maps in Nevada (for now).
It took OSM almost 15 years to evolve toward the current tag for aboriginal lands.
Those editing the map follow the same pattern. In my free time, I follow my own patterns: I focus on geographic areas of interest to me personally, whether it’s places I have lived or places I am visiting. I map hiking paths across the world, while comprehensively am focusing on putting all breweries in my home state of Montana on the map. I’ve been working slowly on tracing the boundaries of Native American reservations in Montana onto the map as well, after Alan McConchie led the effort to improve how their boundaries render on the default base map.
I’ve mapped parts of my favorite ski resorts, sidewalks in neighborhoods I frequent, or the locations of restaurants I love. But rarely am I mapping furniture stores, dog parks, hookah bars, or electric vehicle charging stations. Most of these things aren’t hugely present in my life (for now). I have never mapped in Bolivia, Moldova, Pakistan, or Angola, because I’ve not visited and don’t have connections to these places.
A heatmap of my contributions from Pascal Neis’ “Your OSM Heat Map” tool.
We all map what is relevant to us, as well as try to inspire, petition, or even pay others to map what is relevant to our causes or businesses. Many of us map in places we’ve never visited, and often map based on what data is available. Microsoft’s release of building footprint data in the United States, or Mapillary’s generation of post box and crosswalk locations, means that simple availability of data means more mapping of that data type. Open data in general is a major factor on what gets added to OpenStreetMap, and some data is more proliferate while other types are rare. Because buildings in Beirut have been notorious for having no addresses, a building import doesn’t mean it’s drastically easier to make deliveries or pick-ups there. Meanwhile, a building and address import in Denver means a vast amount of data about not just the buildings but their addresses will affect both the input and output of OpenStreetMap there. In this case, it’s important to consider what data is not available, just as we consider what data is not being mapped by individuals and groups, and what data is not in demand by map consumers.
The OpenStreetMap editing tools also affect the evolution of the project. For many people, a single editing tool—whether it’s OSM iD editor, JOSM, GoMap!! on iOS or Vespucci on Android—is their one portal toward contribution. There are many other ways to contribute. Each editor by necessity must provide the user with certain tools and conveniences, and omit others for the sake of focus or user experience, among other reasons. Just like limited knowledge of French means limited ways to express yourself in French, different OSM editing tools can put blinders on mappers and influence how, where, what, and why they edit. If you’re a Pink Floyd fan, you may be familiar with the timeless lyric:
Run, rabbit, run
Dig that hole, forget the sun
And when at last the work is done
Don't sit down it's time to dig another one.
This is sometimes how I feel when I am tracing buildings with OSM iD editor. It’s a wonderful tool, but the task itself is monotonous, and can drive anybody to madness. On the other hand, the
building_tools plugin in JOSM is a specific tool that helps alleviate these grievances, as explained in Joel Hanson’s recent tutorial. Awareness about the different tools available, across all editors, can really have a significant impact on the direction of OpenStreetMap. More volume of data made available by machine learning means that new data types can be added at scale, where it was missing in the past.
Microsoft building footprints in Billings, MT. Generating the data was automated, but editing OSM will take awhile.
Dealing with huge volumes of data still requires proper tools and proper skills—there are zero cases where machine-generated data has been mass-imported into OSM, without human touch of validation and verification. AI cannot blindly edit the map for us, it can only collect data—sometimes even clean itself for us, like in Mapillary’s verification tools. Will it always be this way? Would you trust an algorithm to not only generate data but to submit automated changesets to OSM? The community position on this stands in stark contrast to many of the self-healing HD map concepts that automotive companies may envision, outside of OpenStreetMap.
Finally, when considering the input and output of the map, we have to admit that most of the map is blank space. The lines, points, and polygons on OpenStreetMap are all instances of data against a background of the unknown. Comparing against a satellite map, we can see from the bird’s eye view that there’s much more on the ground that we have traced and documented. Street-level imagery adds a third dimension, and of course it’s probably good to step away from the computer and step outside to realize the old adage: “the map is not the territory”. I won’t be the 1000th person to quote Borges here, but our map is hardly a 1-for-1 representation of the earth.
We can add curbs, grass lawns, the details of sandy beaches or rocky peaks, every fire hydrant and every line on the pavement. The question in this sense is how much data do we need, how many tagging schemes, to be considered complete? How much is too little, how much is enough, how much is too much? It all depends who is answering the question, who benefits, who edits, and who has the time to really drill down into the details.
Every mapper matters: Tshedy (@Tshedy4) in Lesotho recently completed a 24 hour, one-woman mapathon.
Most importantly, I should repeat that we are all gatekeepers. If you’re editing OSM, you’re affecting the direction of the project. If you’re participating in mailing lists, forums, chat channels, or other community discussions, you have an impact. If you work in the mapping industry, use maps for your daily projects, or develop the tools to edit the map or applications that consume it, you are an important part of this community. The most important thing for all of us is to contribute—whether it is in our professional work, our written and spoken words, or our volunteered map editing.
We need more OpenStreetMap communities, across cities and countries. We need more community leaders. We need to share more knowledge. We need stronger bonds between corporate interests and community interests. We need more people to talk about their personal experiences with the map, and how they are contributing locally. This is a collaborative project, and the only answer to fears about any one collaborater having too much influence is to increase the efforts of the other collaboraters, to be balanced. There is no time better than now to get involved in OpenStreetMap and help contribute to its ongoing growth and evolution.
A snapshot of OSM events in spring 2019, from https://wiki.openstreetmap.org/wiki/Current_events
In the end, what really affects the direction of OSM’s growth? Here’s a short list of what I’ve mentioned here, although it’s doesn’t include the complete picture of reality:
- Who is collecting the data - do they also edit, or even know about OSM?
- Who is editing the map - volunteers and employees
- What data is available - open data, locally or at scale
- The demographics of all people involved - who are they, and who are we?
- Commercial demand - new and old business models
- Humanitarian demand - prevention and response efforts
- Community priorities - what do the locals need?
- Response to political, economic, and environmental events
- Evolution of tagging - a new tag can appear and transform the map
- Editing tools - what do they enable us to do more of, and faster? What barriers exist?
- Data scale - how much detail do we need on the map?
What would you add to this list? Will anything disappear? Will new items emerge?
/Chris, Solutions Engineer