Map Improvement Proposal 10 (MIP-10)

Hivemapper Network
Hivemapper Foundation
6 min readFeb 14, 2024

3/8 UPDATE — After considering the feedback received during the comment period, we have decided to implement a bounty program for app-based contributions. The first weekly bounty will be paid in mid-to-late March, covering Buzz contributions since the previous airdrop on Feb. 14.

SUMMARY — Although the MIP-10 proposal discussed multiple types of app-based contributions, we will start with a bounty for new Buzz reports and incorporate rewards for confirming Buzz reports and providing GPS traces at a later date.

We plan to experiment with multiple methodologies for calculating user rewards in the coming months. As this is an experimental bounty issued at the discretion of the Hivemapper Foundation for development purposes, we do not intend to publish the rewards formulas to prevent gaming.

A baseline bounty will be available for new Buzz reports everywhere in the world. A boosted bounty will be available in 15 areas of the world that have the highest density of dashcams and map coverage, as these regions will have the most viable path toward the density of Buzz reports needed to form a useful product for end customers.

These 15 areas are:

  • Amsterdam
  • Chicago
  • Dubai
  • Los Angeles
  • Lisbon
  • London
  • Madrid
  • Manila
  • New York City
  • Paris
  • San Francisco
  • Singapore (Entire Country)
  • South Korea (Entire Country)
  • Taiwan (Entire Country)
  • Toronto

Background

Although smartphones are not suitable for collecting high-quality map imagery, they are useful for building certain map layers and data products. For this reason, the Hivemapper Network plans to start allowing contributors to contribute data from smartphones for the following uses:

Behavioral Maps

How people drive on a given road segment is often very different from what is legal. In one area, the posted speed limit may be 40 mph, but in reality, traffic typically drives at 50 mph. In another area, the posted speed limit may be 35 mph, but due to the curviness of the road, most drivers travel at an average speed of 25 mph. GPS traces can be used to establish ground truth that is essential for accurately estimating baseline travel times.

Traffic Layers

With a behavioral map of a road already established, GPS traces can also call attention to deviations that indicate the presence of traffic. This is a valuable data layer that is essential for estimating real-time travel times. However, it is important to recognize the limitations of GPS traces. They do not provide information about the cause of traffic. (Is it an accident? Construction? Police activity?) There are other ways to get that information, such as…

Real-Time Reports

Smartphone users can report or confirm safety issues they observe on the road, such as roadwork, road closures, police activity, road debris and more. The Hivemapper Network allows contributors to report these issues to keep the roads safe in a low-distraction way through its Buzz feature. Compared to other real-time reporting apps, the Hivemapper Network has a unique capability to validate reports using community-submitted imagery for the most accurate data. If this MIP is implemented, Buzz reports would be rewarded with HONEY for the first time.

Rewards Model

We propose the following principles for these rewards model:

  • Rewards would vary based on the type of data being contributed and the reputation of the contributor. The more active the contribution mode, the greater the rewards. The greater the value of the individual contribution, the greater the rewards. The more reputable the contributor, the greater the rewards.
  • Across all modes, rewards will be lower than when mapping with a dashcam. One reason is that there is no incremental cost to the contributor. Another reason is that contributions will be passive or completed in a matter of seconds, while mapping with a dashcam requires attention over an extended period of time.
  • Rewards will initially be issued as discretionary Hivemapper Foundation airdrops, enabling flexibility as we learn and make modifications. In the longer term, they would come from the minted rewards pool, following a subsequent MIP process.

The hierarchy of rewards would be as follows:

We anticipate using a similar “contributor score” methodology to the one used for map coverage.

To calculate rewards for contributing via the App, each task would receive a value that could be dynamically determined by factors included in this table below:

During this pilot phase, we expect to limit weekly contributions per contributor. The table below describes our proposed limits.

Please note that these may be adjusted as we gather more data and input from the community.

We anticipate each type of contribution receiving a point value based on the task type and supplemental factors. Every contributor would get a raw weekly score based on their contributions, which would be multiplied by that contributor’s reputation for app contributions to arrive at a final score. Like the reputation for dashcam contributors and AI Trainers, the reputation for app contributors would be between 0 and 1. Any contributor who shows indications of spam or fraud would have their reputation permanently set to 0.

[Raw Contributor Score] * [Reputation] = [Final Contributor Score]

Region Strategy

As with map contributions, we expect to have a weekly reward pool available for app contributors, which would be shared among all contributors based on how much they contribute, as follows.

[Final Contributor Score] / [Sum of All Contributor Scores] * [Global Airdrop Pool] = [Reward]

Based on our learnings from current Hivemapper Network products, we plan to follow a “land and expand” strategy with our app-based products to reach critical mass rather than offering them worldwide at launch.

Initially, rewards will only be available in certain regions to achieve critical mass. We anticipate offering this feature in approximately 15 metropolitan areas at launch. Additional regions would be added over time.

As this contribution mode achieves scale, another MIP would be presented at a later date to transition the rewards from a special Foundation Reward to the weekly rewards pool.

Feedback on the proposal

Members of the Hivemapper community shared many useful questions, suggestions, and comments in the #mip-10 channel of Hivemapper’s Discord server. This shaped our thinking around the bounty, especially relating to spam mitigation and the region strategy. To summarize some of the questions we received:

How will you prevent users from spamming Buzz? Let’s be honest: whenever a token reward is available, people inevitably try to game it. It was true for dashcam contributions, it was true for AI Trainers, and it will be true for app-based contributions as well. Over the past year, we built an ever-more sophisticated anti-spam program for AI Trainers that makes it extremely difficult to submit spam. Likewise, we made extensive improvements to the rewards formulas for map contributors to align rewards with the needs of the map. We will do the same for Buzz. The most common ways bad actors will try to spam Buzz are fairly obvious, and we have put measures in place to mitigate poor or duplicative data. However, we believe the best way to build is based on real-world learnings, rather than trying to design a perfect solution from day one.

Will this data actually be useful yet, with a low volume of reports? This is a fair point. Of course, you need to reach critical mass to make any network useful. It is a cold start problem, a chicken and egg problem, however you put it. We have shown that a bounty is an effective way to overcome this and kickstart a new product — just like we did with AI Trainers last year. This is why we are focusing the bounty on the regions of the world with the strongest dashcam penetration, rather than issuing the same reward everywhere in the world.

If a user submits a Buzz report in an area without dashcams, how will you confirm it? Legacy products such as Waze face the same challenge. They have built extensive algorithms to confirm reports based on factors such as the number of reports from credible submitters. We can do the same — but with the inherent advantage of having ground-truth data from dashcams to help us confirm the credibility of Buzz submitters where available.

--

--

Hivemapper Foundation
Hivemapper Foundation

Published in Hivemapper Foundation

The Hivemapper Foundation is a not-for-profit organization formed in 2022 and based in the Cayman Islands. Over time, the foundation will take on many of the responsibilities currently held by Hivemapper Inc.

No responses yet