What are the relevant research areas and/or research questions for VGI/Crowdsourcing in the future ?

Rapporteur: Peter Mooney

Source: Flickr image - flipchart page 1
Consider the A⁴C⁴ discussed by SwissTopo:

• Authority
• Accuracy
• Availability
• Actuality

These can be considered as part of the "system". However when we consider the "C" - these are more difficult

• Completeness: partial sampling is not good. It is important to look at the "mapping curve" and see what has been done or what is being done. Where can you actually trust this graph?
• Coverage
• Consistency: when representation is not uniform over space, where only zones of consistency exist
• Correctness

Could we teach machine learning algorithms to consider some of the "C" parts? The outputs from machine learning could then be verified by people against the actual mapping work carried out. However this is not an easy task. We will need: sufficient amounts of data, to trust the people doing the mapping, to be able to measure or ensure consistency, and allow for multiple languages in our machine learning algorithms.

Source: Flickr image - flipchart page 2

How do you interact/train "social processes"

Opportunities here for input from experts in human factors. How do we go about resolving disputes in crowdsourced mapping? What about the signal to noise ratio of the data which is generated?

Return on Investment (ROI) in Crowdsourcing of Spatial Data

The best ROI may come from the long tail or peak contributors to a project. With these contributors you might get as good an ROI as involving a very large number of contributors. A small number of very dedicated people can be very effective. But this will always depend on the task(s). Depending on the task(s) it might be necessary to involve large numbers of contributors.
Perhaps the best approach might be to consider small tasks/challenges with some type of reward system in place. This could encourage contributors to map or collect something "new" or "rare" or to "go somewhere new" to map.

Where does Remote Sensing/Satellite Imagery have a role to play? Consider the potential of a daily return of one satellite every day on a continuous basis. There will be new training challenges. However it could also allow for new forms of feedback - VGI could verify the classification of an image (for example)

Source: Flickr image - flipchart page 3

Dicussion of a common feedback chain or channel between NMCAs and VGI

Feedback processes do exist within private companies. Feedback is also part of the process in VGI and NMCAs.
Semantics is a common issue between the two sides
VGI and NMCA are actually trying to understand and represent the same "concepts".
This shared problem opens up the opportunity to work around license problems. Errors do not have any particular license! Errors can be considered as shared work

Is the research focus on VGI too narrow at the moment? One could argue that there is a North-West VGI - but is this representative of a Global VGI?
Which application areas or policy areas could be addressed to deliver more innovative solutions if we compliment or combine NMCAs and VGI?
This is a very interesting question - but not such an easy one to answer.

Looking forward - we need to find use-cases where VGI and NMCAs can work together on NEW problems. Think about this siutation with the percieved boundaries of NMCAs. Think about the roles of the NMCAs.

Source: Flickr image - flipchart page 4

Use-cases - Shared NMCA and VGI problems

Rather than considering places where NMCAs and VGI are different - are there use-cases where both NMCA and VGI could work together? These use-cases would be of mutual benefit to both sides while at the same time addressing issues which are of value to wider society, government, etc. The following is a listing of the ideas which were suggested - lots of interesting food for thought here - also lots of interesting starting points for collaboration and research.

1. End-to-end journey patterns, multimodal transportation systems, multimodal journies
2. Demographics. Who maps? What do they do? Is the profile the same within the VGI projects and the NMCAs? What are the common attributes?
3. Fluxes - social fluxes, urban fluxes
4. Population: We have things like very good building descriptions but not very good information about the population inside those buildings. Think about a night vrs day time census idea.
5. Urban Forms and Functions - some cities have actually no data on this. But VGI could actually supply something here!
6. Resilence: Response to emergencies - this is something that will effect everyone. Shared work on 'how' we respond.
7. Agriculture: Issues such as precipitation coverage
8. Place vrs Location - informal geographies
9. Points of Interest, neighborhood scales
10. Visualisation - both sides understand the value of good visualisation, 3D, moving processes, multidimensional, etc
11. Missing Buildings, Demolished Buildings, Illegal settlements. What are the consequences for illegal settlements in flood prone areas for example?

What are the relevant research areas and/or research questions for VGI/Crowdsourcing in the future ?

Rapporteur: Anthony Simonofski and Joep Crompvoets

We decide to separate the general question into two perspectives : NMCA and non-NMCA

1: NMCA

• How to generate savings from the VGI ?
• How to provide up-to-date data and correctness of data ?
• How to involve the crowd in new products ? In 15 years, the crowd won’t read any maps and will only need information at the right place. These changing models must be taken into account.
• We should divide the VGI practitioners in two categories : kids (education), professionals, citizens and interest groups (hikers,…). It would be nice to have a labelling of the data quality. Ex: for the professionals, 80% quality is sufficient. There is a connection with the “Fit-for-purpose” quality. What are the different labelling of data quality according to the categories of practitioners ?
• How to validate (automatically) the data that is generated by VGI ?
• How NMCA can efficiently exploit the timing capabilities of VGI ? (Maybe in real-time)
• We must distinguish the production cycles of data and the integration of core datasets with the VGI.
• Another key problem is to sustain engagement. We have to work with social psychologists to find out what drives the practitioners to contribute ? What are the cognitive triggers of engagement ?
• We also have to distinguish the new data for existing layer or new data for new layer. How to manage the processing of data and distribute it into new or existing layers ? This is linked with the quality of data with different validation mechanisms in function of the layer.
• Transitional mapping : how can we change the maps and go back to other sorts of maps ?
• There is a need for guidelines about the ownership of VGI Data. Furthermore, the specific question of “How to handle personal data” is also essential. (ex: images of people on photos).
• How to motivate the government to be part of the crowd ? It is relevant for national or local datasets.
• Are the ISO standards applicable of data quality assessment of VGI ? There is a need for external measures. You have two models : intrinsically motivated crowd and externally payed crowd?
• How can VGI be impacted by standards ?
• How to encourage people to use NMCA data/systems instead of other systems ? Adoption of the technology in the hand of the public.
• How to come with more generic methods to adapt to different types/applications of VGI ? Too many papers are suggesting conceptual papers but don’t address anyone’s needs.
• Raising quality of data is a problem for land cadasters. VGI is a contributor to improve the data. In the Netherlands and Finland, they are working with landmarks between parcels and sensors to sense the quality of measurement.
• We should have a taxonomy of stakeholders that are interested in VGI and have different motivations depending of the category of stakeholders (and thus take actions accordingly). How can we continue to motivate the “crazy-mappers” after all empty spaces are filled ?

2: Non-NMCA

• The difference between NMCA and Non-NMCA is that NMCA have regulations that forces them to process a set of data. Depending on the layer, the influence of the law is different.
• “Médecins sans frontières” : working between official maps and other sources. This is also related to transitional mapping
• Identify which crowdsourced data can calibrate the algorithms. In that regard, we don’t need high quality data or high volume but we need access to specific information.
• Identify which metadata is necessary for which data
• Composition and dynamics of national, building stocks. They have little information about age, history and morphology of buildings because they are owned by the tax office.
• They also have to make the maps available online.
• Using VGI is making the parallel between administrations that do not want to open their data and VGI that provides data of high quality about similar areas.
• We also need a platform to share and integrate the appropriate data for the determined question. This platform can be used to refine the algorithm before an emergency happens.
• How would we increase resilience and minimizing the risk using VGI ? (ex of application : risk management).
• How can we represent the current geo-localized data to the crowd (that are not specifically professionals)? Possible lead: switch from 2D to 3D. What is the ideal representation/facility management for each situation? (ex: for blind peoples ?)
• VGI in emergency : Would it be possible to share geo-spatial data from drones ? (a platform already exists : dronesmap. It is a wiki-based product for general purpose).
• The overlap between using social media, collect data, and VGI. The interaction between these interests are interesting to tackle in the future.