Cybersyn Capitalism & Algorithmic Regulation
The New Yorker has a fascinating essay by Evgeny Morozov on Project Cybersyn: the Chilean socialist government of the 1970’s attempt to centrally direct the nation’s industry. This sought to marry cybernetics — the study of communication and feedback loops within biological, technical and social systems — and cutting-edge technology to create a system of socialist planning which also gathered local knowledge from the factory floor.
The project’s Operation Room speaks volumes of their ambition. Screens lined the hexagonal room’s walls, and tables and paper were banned. One armrest on the orange and white swivel chairs was lined with buttons to navigate the screens, whilst the other featured a drinks holder and ashtray. Historical and real-time production data was to be broadcast onto the walls, with simulations to model the effect of centrally-planned decisions. There was even ambition to measure the real-time happiness of citizens via a mood dial in their living rooms.
The project was an impressive and hyper-futuristic attempt to overcome Hayek’s knowledge problem. Unsurprisingly, the scheme demanded far too much of the technology of the era and was plagued by bureaucratization. It was overthrown along with Chile’s socialist government less than a year after its launch.
Fast-forward 40 years and the private sector has raced ahead with cybernetic insights and the efficient exploitation of information. As Morozov points out, our Google-owned Nest thermostat knows not just to adjust temperatures but whether we’re home or not. Uber tracks the GPS of its drivers and customers and uses an algorithm to determine ‘surge pricing’ at times of high demand. Amazon has a patent on ‘anticipatory shipping’ to deliver you something you haven’t even ordered yet. The information companies hold on us can yield intriguing insights — Google searches can predict flu outbreaks, and Target can work out when you are pregnant. Regardless of the limitations of such data-crunching, this is Cybersyn Capitalism — and the network of interconnected devices monitoring, tracking and relaying real-time information on us is only going to grow as the so-called ‘Internet of Things’ comes into its own.
Governments are raring to harness such tools, technology and insights for themselves. There’s ‘big data’, ‘smart data’ and ‘smart cities’, and the collaborations with corporations to make it happen. Such efforts are politically uncontroversial; cross-party, innovative, efficient and ‘nice’.
‘Algorithmic regulation’ is a concept popularized by ‘the oracle of Silicon Valley’ (link) Tim O’Reilly which explicitly seeks to blend the private sector’s cybernetic approach to data with a new era of governance.
Regulation by algorithm is what successfully keeps spam out of inboxes, ranks search results and detects credit card fraud. Such regulation, O’Reilly notes, features a ‘deep understanding’ of the desired outcome, real-time measurement to check if it is being achieved, algorithms to make adjustments based on new data, and periodic review of the algorithms’ effectiveness. This in contrast to the state’s typical regulatory models, he argues, which are focused upon clunky and often outdated rules, rather than on outcomes. Instead, legal regulations should be considered ‘a constantly updated toolset’, adjusting according to data and what is shown to be working. O’Reilly encourages government to emulate the ‘lean startup’ approach, where each business is considered a ‘machine for learning’ and discovery.
What could be wrong with Silicon Valley-ifying government and sharing capitalism’s cybernetic insights with the state? If done effectively there’s no doubt that it could make government machinery more efficient, streamlined and discreet. Government by algorithm focused upon ‘what works’ also sounds appealingly anti-political: what Morozov brands in another thought-provoking essay on big data as ‘politics without politics’.
Rule by algorithm, it seems, can take the noise and ideological bias out of economic decisions, and replace the rhetoric of politicians with cool, clean, data. There’s a certain technocratic appeal to it: we know that voters are often wildly misinformed when it comes to political fact. If only we could engineer policies (say, the supply of visas to the UK) which are responsive and based on the world as it is, instead of the world as voters perceive it!
And yet, just as algorithmic government may carry some technocratic appeal, it also leans towards a more dystopic technocracy. Algorithmic regulation is still regulation, and has the potential to be more pernicious and far-reaching in its effects than conventional forms.
O’Reilly notes that ‘new technologies make it possible to reduce the amount of regulation, whilst increasing the amount of oversight and desirable outcomes’. He makes this point to demonstrate that reputational systems like those used in AirBnB lessen the requirement of government regulation, but history shows that when faced with the possibility to expand their areas of operation, governments are only too happy to do so.
And, as we have more data to hand and develop smarter ways to exploit it, the potential scope of government grows. A ‘smart government’ might not just have the power to provide effective services, but to adjust the behavior of its citizens, whether that’s through the use of smart parking meters or something more insidious. Behavioral economics and the ‘libertarian paternalism’ of the nudge economics beloved by Cameron thrive off the use of such data. And there certainly isn’t a political reticence in going down this route. Morozov points to two think tank studies call for linking tax rebates and benefits sanctions to a citizen’s activity level. Recently, the Conservatives unveiled plans for a benefits smartcard to be used only on ‘essential’ items. Such measures channel the spirit and capabilities of a social algorithmic regulation, and further extend the use and creation of data on citizens.
Silicon Valley is by no means a libertarian stronghold, and (for all their talk of technology empowering and unshackling us from constraints) it’s a mistake to assume that a tech buff’s policy ideas would necessarily liberate individuals in a political sense. For example, Tim O’Reilly suggests a use of algorithmic regulation of using GPS to automatically ticket speeding drivers, readily admitting that ‘most people today would find that intrusive and alarming’.
Whilst Chile’s Cybersyn project was predominantly concerned with a co-coordinated economy, today’s cybernetic bureaucrat has the ability to delve much deeper into the workings and behavior of society.
It’s also important to note that the use of big data or algorithmic tricks in government activity doesn’t mean somehow make a function less political. The use of technology in performing an action does not make it technical rather than a political one. Choices made by a ‘smart’ government are still political (no matter how much grunt work is delegated to a computer model) —for they are fundamentally based upon some political preconception of a desirable outcome to be achieved. An algorithm to determine the optimum level of visas, for example, depends on a political assumption of what the optimum level of immigration or type of immigrant should be. Choices to ease the load on emergency service or encourage healthier eating (and the ways to do this) are similarly political, even if not as obviously so as using face recognition software in CCTVs to detect suspicious behavior.
Moreover, constructing any algorithm necessarily involves decisions of what is relevant, significant, and a desired outcome. Even outside the realm of government these issues are often political in a broader sense.
For example, Google adjusts its search algorithm up to 600 times a year, and when doing so can make controversial decisions on how to provide the best search result, promote its own services over those of competitors, and facilitate the demands of media companies (eg, by demoting search results of sites which persistently violate copyright). Bitcoin is sometimes referred to as a ‘non-political’ monetary unit because of its detachment from the control of central banks and its operation by open source code — yet its very creation was an deliberate refutation of the status quo, with politics written into its very code with its pseudo-anonymous nature and intentionally fixed supply. Politics by algorithm might seem to take the politics and tough choices out of decision making, but we should not forget the fact that decisions have indeed been taken, and accountability for these will lie with a human. Otherwise, what happens when the computer says ‘no’?
There’s plenty of things to disagree with Evgeny Morozov on (not least his stance on private sector innovation), but he’s right to flag up the potential creep of state technocratic power which comes from assuming that technology can take the difficult choices out of politics.
This doesn’t mean for one second that governments shouldn’t use particular insights to improve government services or regulations, nor does it necessarily make it objectionable for them to store and generate data on the population. (Such activities also involve wider questions of privacy, utility, identity and the ethical use of data.)
It does mean that it’s important to remember that the more useful information government holds, the greater their temptation to exert more influence over people’s lifestyles and economic activity. We should also remember that even if an action seems apolitical because of the technology involved, there can still be political choices behind it requiring assessment and debate. Tech does wonderful stuff, but we shouldn’t be complacent in expecting it to find answers to political questions. If we do, we’ll find the assumptions are already smuggled into the algorithms.
theartofmadeline
yousharknotpass
titsay
izzy806
