I wonder how we can implement the social contract via technology, and how that can be done democratically.
A case study to explain what I mean…
One of the slow controversies in London over the past year has been the Low Traffic Neighbourhoods (LTN) programme: closing many residential streets to road traffic, sending cars onto main roads instead. There’s some background here including how it was built out of a schools-focused programme during the first lockdown (streets outside schools and on regular walking routes were closed to cars).
LTNs are a joy and a pain.
The future of the city involves fewer cars, we all know that. Walking on these quiet streets and having coffee in the parklets now built outside cafes is transformative. BUT the schemes channel cars onto already congested main roads, and semi-local trips that aren’t well served by buses are made much more difficult.
Jimmy Tidey’s brilliant research has shown how LTNs kicked off a culture war on Twitter – though catalysed by a relatively small number of vocal black cab drivers. There are posters in almost every local shop against LTNs and they’re often vitriolic. I spotted a banner headline, All Streets Matter.
Breathtakingly tin-eared.
For me, the root of the vitriol is that two constituencies of people feel it is unfair.
- Local people feel like these are our streets – why shouldn’t we be able to drive down them?
- Black cab drivers used to have special access to The Knowledge: a detailed mental map of London’s short-cuts, effective precisely because it was specialist. But it was made accessible to everyone with Google Maps (which, I remember hearing, has contributed to a 30% traffic rise on residential streets over the past decade). And now we’re being “punished” for the traffic with street closures, but the cab drivers feel it’s not their fault and they should still have access.
The problem is exacerbated by technology. The LTNs are often in effect for only some of the day, so the street isn’t physical blocked. The closure is implemented by a road sign, cameras with automatic number plate recognition, and penalty fines sent through the post. One of my neighbours has been stung by a series of 65 quid fines, having sailed through computer-closed streets accidentally a number of times. So, poor software.
But technology is also perhaps part of the solution!
Long term we’ll have self-driving cars. We won’t need to close streets with bollards and impose fines – the cars can be programmed. The Low Traffic Neighbourhood policy will be a software point release.
So let’s think about how to bias that future pathfinding algorithm for fairness.
Perhaps what we’ve identified is that local people have more “moral right” to use their neighbourhood roads than people from across town who are using the street as a shortcut. Those people from across town feel like freeloaders: they’re taking the benefit of the cut-through but they don’t have to live here.
(Something similar happened in Los Angeles when Waze became popular. People went to extraordinary lengths to protect their local streets by fooling the Waze maps. As discussed here in December 2020.)
Could we say that fairness means: local streets for local people?
What if we had some way of categorising roads on a spectrum from small (local and residential) to big (thoroughfares)? If you live within 1 mile of a small road, it’s free to use. Over a mile, it’s thoroughfares only or you get a penalty.
The existing Low Traffic Neighbourhoods would see some cars again, but traffic volumes would be low: the streets would be closed to any car from outside the local area.
Ok, as a thought experiment that works for the future. NOW we can ask about how to implement this without waiting for robot cars. Could LTNs be implemented in software today?
From a product perspective, the answer is yes.
Let’s imagine we have multiple routing modes in Google Maps. Perhaps the different algorithms are embodied as different characters, just like each ghost in Pac-Man embodies a different search algorithm. (I’m picking on Google Maps but I’m using this as a stand-in for all routing apps.)
In addition to the “quickest” mode, and the “most fuel efficient” mode, there would be “social contract mode” – which would be the default. This mode would avoid residential streets outside a 1 mile radius of your home or 0.5 miles from your destination. And it would be the default.
Through legislation, “social contract mode” for map routing would be mandated for all in-car navigation from 2026.
Sounds plausible. The question then becomes… how could a policy like this get enacted? Three challenges:
- Politicians and civil servants don’t know enough about technology to know what is hard and what is easy. Nor the media or the public: technology is a specialist subject. So how can we have a meaningful public debate – which is what we need for democracy?
- Mandating a product solution is the wrong level of abstraction: but let’s say the outcome was somehow made law, how would that even be expressed? How is it possible to express a requirement like this in law? The social contract must be baked into product features?
- Once we open the door to the state interfering in tech product decisions, how are stupid decisions defended against? What about dangerous decisions, or ones that reduce liberty? We have norms and laws and centuries of philosophy (and in some countries, a constitution) for the limits of how the state may limit the freedom of the individual, but we aren’t nearly as sophisticated when it comes to technology.
I don’t know the answers to these, but the utility of having a specific case study such as Low Traffic Neighbourhoods is that we have something concrete to debate.
London traffic is a specific case of something general and important, which is how society uses technology to enact its values, and what the mechanisms and limits on this should be.
Another instance of the general problem is Facebook’s engagement algorithm. Can society really tell Facebook how to tune its systems for chasing engagement, given that the ad-supported model requires it? Can we really insist that Facebook puts a cap on engagement, reducing its profit margins, or even changes its business model to include paid services – which will reduce accessibility?
I mean, yes we can and should be having that debate. The extremism caused by Facebook’s algorithms can be seen as a public health problem and, if that analogy holds, I can point out that we’re perfectly happy to tax the cigarette companies without outright banning them. (Paying for externalities is one of the uses of tax.) So maybe the same approach should be adopted with Facebook.
But the question is the same: how should the desired social outcome be expressed as a technology product requirement, and how can it be expressed in law?
There are social values baked into software already. We need democratic ways to tune the parameters.
I wonder how we can implement the social contract via technology, and how that can be done democratically.
A case study to explain what I mean…
One of the slow controversies in London over the past year has been the Low Traffic Neighbourhoods (LTN) programme: closing many residential streets to road traffic, sending cars onto main roads instead. There’s some background here including how it was built out of a schools-focused programme during the first lockdown (streets outside schools and on regular walking routes were closed to cars).
LTNs are a joy and a pain.
The future of the city involves fewer cars, we all know that. Walking on these quiet streets and having coffee in the parklets now built outside cafes is transformative. BUT the schemes channel cars onto already congested main roads, and semi-local trips that aren’t well served by buses are made much more difficult.
Jimmy Tidey’s brilliant research has shown how LTNs kicked off a culture war on Twitter – though catalysed by a relatively small number of vocal black cab drivers. There are posters in almost every local shop against LTNs and they’re often vitriolic. I spotted a banner headline, Breathtakingly tin-eared.
For me, the root of the vitriol is that two constituencies of people feel it is unfair.
The problem is exacerbated by technology. The LTNs are often in effect for only some of the day, so the street isn’t physical blocked. The closure is implemented by a road sign, cameras with automatic number plate recognition, and penalty fines sent through the post. One of my neighbours has been stung by a series of 65 quid fines, having sailed through computer-closed streets accidentally a number of times. So, poor software.
But technology is also perhaps part of the solution!
Long term we’ll have self-driving cars. We won’t need to close streets with bollards and impose fines – the cars can be programmed. The Low Traffic Neighbourhood policy will be a software point release.
So let’s think about how to bias that future pathfinding algorithm for fairness.
Perhaps what we’ve identified is that local people have more “moral right” to use their neighbourhood roads than people from across town who are using the street as a shortcut. Those people from across town feel like freeloaders: they’re taking the benefit of the cut-through but they don’t have to live here.
(Something similar happened in Los Angeles when Waze became popular. People went to extraordinary lengths to protect their local streets by fooling the Waze maps. As discussed here in December 2020.)
Could we say that fairness means: local streets for local people?
What if we had some way of categorising roads on a spectrum from small (local and residential) to big (thoroughfares)? If you live within 1 mile of a small road, it’s free to use. Over a mile, it’s thoroughfares only or you get a penalty.
The existing Low Traffic Neighbourhoods would see some cars again, but traffic volumes would be low: the streets would be closed to any car from outside the local area.
Ok, as a thought experiment that works for the future. NOW we can ask about how to implement this without waiting for robot cars. Could LTNs be implemented in software today?
From a product perspective, the answer is yes.
Let’s imagine we have multiple routing modes in Google Maps. Perhaps the different algorithms are embodied as different characters, just like each ghost in Pac-Man embodies a different search algorithm. (I’m picking on Google Maps but I’m using this as a stand-in for all routing apps.)
In addition to the “quickest” mode, and the “most fuel efficient” mode, there would be “social contract mode” – which would be the default. This mode would avoid residential streets outside a 1 mile radius of your home or 0.5 miles from your destination. And it would be the default.
Through legislation, “social contract mode” for map routing would be mandated for all in-car navigation from 2026.
Sounds plausible. The question then becomes… how could a policy like this get enacted? Three challenges:
I don’t know the answers to these, but the utility of having a specific case study such as Low Traffic Neighbourhoods is that we have something concrete to debate.
London traffic is a specific case of something general and important, which is how society uses technology to enact its values, and what the mechanisms and limits on this should be.
Another instance of the general problem is Facebook’s engagement algorithm. Can society really tell Facebook how to tune its systems for chasing engagement, given that the ad-supported model requires it? Can we really insist that Facebook puts a cap on engagement, reducing its profit margins, or even changes its business model to include paid services – which will reduce accessibility?
I mean, yes we can and should be having that debate. The extremism caused by Facebook’s algorithms can be seen as a public health problem and, if that analogy holds, I can point out that we’re perfectly happy to tax the cigarette companies without outright banning them. (Paying for externalities is one of the uses of tax.) So maybe the same approach should be adopted with Facebook.
But the question is the same: how should the desired social outcome be expressed as a technology product requirement, and how can it be expressed in law?
There are social values baked into software already. We need democratic ways to tune the parameters.