Stories
Slash Boxes
Comments

SoylentNews is people

posted by cmn32480 on Saturday September 24 2016, @03:51AM   Printer-friendly
from the where-to-make-the-dump dept.

A functioning society requires various public services, such as hospitals, schools, landfills, etc., but deciding where to build them can be a complicated and contentious issue. The cost to build them are typically shared across society in the form of taxes, and deciding where they go involves various optimizations and tradeoffs to maximize their impact on society while minimizing the cost to society. Optimizing this kind of decision making is an active research topic in the field of algorithmic and network game theory.

A very common approach is to address the issue from a centralized, top-down perspective whereby a city planner considers the network of society as a whole, and inputs the pros and cons into a global optimization algorithm to find a minimum cost solution. This approach is known to not provide the most optimal solution. Another approach is to let individual agents make decisions in best response to the choices of their neighbors. For instance, if a region can access a hospital in a neighboring region, then they would have little motivation to want to have a hospital built in their region; however if none of their neighbors have a hospital, then they would be more likely to be willing to have one built in their region. Although this sounds more appealing than the top-down approach, this approach is also known to not be very socially efficient.

Yi-Fan Sun and Hai-Jun Zhou from the Chinese Academy of Sciences have published an open access paper in Nature's Scientific Reports that considers a cooperative decision process where global decisions are made via local consensus.

Briefly speaking, the basic rules are that agents in need of service recommend their network neighbors of highest local impact (to be precisely defined later) as candidate service providers, and an agent may be chosen as a service provider only if all its non-server neighbors are happy with this appointment. This distributed selection mechanism does not require the global structural information of the system but only involves local-scale information exchange. Yet very encouragingly we find that it leads to socially efficient solutions with tax level approaching the lowest possible value.

[Continues...]

In their idealized model they only considered construction costs to build the facilities and not other imposed societal effects such as changes in traffic patterns or local environment impacts. It is hard to say how well this approach would work in practice, but it would be interesting to see whether NIMBYism is borne out of a true opposition to an issue, or whether it is more rooted in the fact that centralized decisions get imposed upon locals without much consideration of their input. Regardless of whether this approach would work with humans, it is directly relevant to AI research such as robot swarms for finding efficient mechanisms for global decision making.

From the theoretical point of view, the demonstrated excellent performance of the local-consensus mechanism is very encouraging. Our work suggests that it is theoretically possible to efficiently solve the service location problem by distributed decision-making. The local-consensus mechanism does not need a central planner and it does not require the structural knowledge about the whole network. Furthermore, every agent participates in the decision-making process and its opinion has been incorporated in the final cooperative solution, which may help stabilizing the solution.

From the paper's abstract:

We discuss the issue of distributed and cooperative decision-making in a network game of public service location. Each node of the network can decide to host a certain public service incurring in a construction cost and serving all the neighboring nodes and itself. A pure consumer node has to pay a tax, and the collected tax is evenly distributed to all the hosting nodes to remedy their construction costs. If all nodes make individual best-response decisions, the system gets trapped in an inefficient situation of high tax level. Here we introduce a decentralized local-consensus selection mechanism which requires nodes to recommend their neighbors of highest local impact as candidate servers, and a node may become a server only if all its non-server neighbors give their assent. We demonstrate that although this mechanism involves only information exchange among neighboring nodes, it leads to socially efficient solutions with tax level approaching the lowest possible value. Our results may help in understanding and improving collective problem-solving in various networked social and robotic systems.


Original Submission

 
This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 1) by Francis on Saturday September 24 2016, @10:25PM

    by Francis (5544) on Saturday September 24 2016, @10:25PM (#406057)

    It doesn't even need to be unanimous, just requiring consent causes all sorts of trouble, both with NIMBYs as well as people who demand desirable services be placed in their neighborhoods.

    The local school board has meetings with the public when they decide what schools to close and where to locate new schools. It's been an unmitigated disaster. The areas where the parents have the most time and money tend not to see their schools closed and they tend to see them opened earlier than areas where it would make more sense.

    The net result is that the children from better off parents wind up closer to schools than the students in poorer areas. Not to mention that the people doing the screaming now aren't necessarily the same people that are going to need services in the future which tends to be more for the children coming from low income families.