Today i am #WorkingOutLoud sharing the second full draft of the chapter on ‘Algorithmic Wars’ from ‘The Social Age Guidebook’. I feel this second draft is stronger, but still needs work.
Humans are pattern recognition machines, so it’s ironic that we are facing such a struggle conceptualising, and coming to terms with, machines that can determine the patterns of humans.
And yet that is exactly the foundation of the Algorithmic Wars we face: the new battleground of sense making, productivity, and power, in the Social Age.
It’s partly a battleground of ignorance and misunderstanding, partly a battle about existing power and disruption, and partly an exploration of what it means to be a self determining, free, ’human’, and how free we really are.
In many ways it’s a war that will determine our future as a species: if we harness the power of machine learning and artificial intelligence in ways that are broadly productive, highly creative, and equitably owned, then we may find a society that is more equal and better off at scale. If, however, we leave it to the market to determine the outcome, we will doubtless end up more efficient, but more exploitative, and less equal than we are now.
When i was twelve, and busy trying to fail at maths, i got as far as ‘equations’. And ‘algebra’. Teachers would occasionally set challenges, typically involving cars, distance, destinations, and efficiency, and i would struggle to discern meaning from chaos. At no point did they ask me to mathematically predict human behaviour, or wrest control of policing, or policy, from human hands.
Math was largely the purgatory between ‘History’, and ‘Home’, best endured sat next to the window, which facilitated daydreaming.
Today, more than at any point in human history, ‘maths’, characterised and expressed as ‘algorithms’, rules our lives.
It keeps planes in the air (or fails to do so for inexplicable reasons), minimises the amount of time it takes an ambulance to reach you, determines the price of your wheat and gas, and directly impacts on the words that your politicians speak to you when campaigning.
Not specifically because the Organisations behind these things have recruited people who were brilliant at algebra, but because computing power, and the conceptual frameworks of programming and analysis that it enables, have evolved.
The tools are now more powerful than the hands that wield them.
When we hear conversations about ‘algorithms’, we are typically not simply hearing about hard problems that are solved faster by computers: we are hearing about radically large and hard datasets that are being ‘understood’ by computers.
If you run a warehouse distribution business, then knowing where all of your stock is sitting, what condition the tyres of your trucks are in, and how many hours overtime you are paying, are all hard problems to solve, but can all be solved in predictable ways. Machine learning may help optimise the systems that do this.
But the real power comes from prediction, based upon large data sets: if you can optimise where to keep your stock and dynamically adapt pricing to match stock levels, if you can correlate tyre wear to individual driver behaviour, and track overtime back to optimisation of stock holding, then you can reframe your business to be both more effective and more cost efficient. It’s this second type of understanding that we are attacking, and the questions will be around who controls it, who benefits from it, and who loses out. Like the driver who is fired for braking too heavily.
We are seeing pattern recognition at scale, predictive power unleashed, and a level of understanding that would be beyond us as humans, no matter how engaged i had been in that class.
It’s a search for the shadows and whispers of patterns that exist in the footprints we leave.
‘Algorithms’, in the contemporary context of debate, are radically complex predictive, and analytic, systems, which enable us to make sense of large scale data at speeds that are typically useful.
I would venture that if we had to characterise the foundations of Algorithmic War, it’s not the specific outcomes in isolation that are usually the issue (although in some cases, most certainly are the issue), but rather it’s the broader context of those outcomes, and the ways that those outcomes become inescapable, as we feel the imposition of new systems of organisation, sense making, and power, at great scale, and speed.
It’s the way that algorithms give rise to new types of power, and how that power is impacting back into our wider society.
Take Facebook (an easy target, i realise, but when ‘sense making’, it’s ok to start at ‘easy’). Contemporary criticism of Facebook hinges on how the hidden algorithms give us something unexpected, undesired, or somehow deceptive: by ‘choosing’ one news item over another, by ordering and regulating my ‘feed’, by filtering future stories based on my profile and interaction with current stories, we find ourselves individually, and collectively, in a new space. We like to think that we make sense of the world by looking around us, gatherings news, evidence, opinion, and fact, and making a judgement. We react badly if we feel those inputs are being deliberately skewed.
And yet, of course, we have never been the objective problem solvers that we would like to think we were: every way we look at the world is through a filter, and the context of stories is personal in every case. But despite these failings, we have at least felt an element of control: i can ‘choose’ to read the Guardian, or the Daily Mail, i can watch Fox or the BBC. I can choose who is in my community, and who is outside of it.
Those people who take issue with Facebook may be described in two camps: those who feel that the ‘well intentioned’ algorithms are driving undesirable outcomes (echo chambers, inappropriate juxtaposition of content, filtering out of alternative views etc), and those who feel that the fundamental technology is bad, and possibly being used in deliberately deceptive ways (fake news, interference in democracy, creation of artificial social movements, and pseudo viral effects).
Or to put it another way: in one view, we are progressing in broadly positive ways, but with highly undesirable side effects and consequences, or we are progressing in fundamentally flawed ways, deceived by technocrats who are unaccountable to anyone.
This is reflected in the responses of the wider system: governments seek to regulate, to control poor effects, whilst concurrently seeking to automate, to maximise beneficial ones.
Individuals seek to disengage and tune out, to minimise concerns on privacy, and deliberate bias, whilst seeking to maximise individual gain (through optimising individual utility and personal value), and amplifying those messages that mean the most to us within our local tribe.
Predictably, we are in a conflicted time, hence that term ‘Algorithmic War’, because it’s not an outright acceptance, or rejection, of the technology that is at stake: it’s more about how we can evolve our structures of understanding, and effect, in considered ways.
Because one thing is certain: if we do nothing, then technology will take us into places that we, as society, are entirely unprepared to go. And we are already well down that path.
In popular media, in Organisational adoption, and in the initial narratives of success, or failure, we are often acting as unconscionably naive, or unhelpfully vague. For example, we understand that ‘bias’ can be an issue, but that leads to populist narratives around inherent bias that simply do not stand up to scrutiny, for two reasons: firstly, that there is nothing inherently biased about every machine learning system, there is just bias in the data we feed them, or sometimes bad design, and secondly, a failure to realise that a core feature of machine learning systems is that they learn.
So how things are now is not how things will always be.
This was my conversation with a taxi driver in London: he accepted the arrival of self driving cars, but described how they would not know how to react to a pigeon in the road. He said that ‘professional’ drivers knew to just keep driving, because pigeons always took off at the last minute. So he could accept that self driving cars could learn to drive, but could not accept that they could equally model the pigeon avoidance mechanisms of professional drivers, and learn to do that too.
Some issues are more clearly emergent ethical conversations: should we feed people up images of self harm and suicide (some evidence shows that the ability to explore these topics can lead to better outcomes), or is it simply exacerbating the issue.
Should we have adverts for McDonalds showing up next to those posts as well, or should they be held more ‘respectfully’. As a society, i have no doubt that we will figure these things out, although not without some fails along the way.
Because these questions are not about quantifiable success: they are about ethics, about belief, about the reinforcement of dominant narratives, and about self righteous indignation, as well as about justifiable self righteous anger. Sometimes all those things at the same time.
These emergent and ethical conversations (about privacy, about decency, about protection, safeguarding, and harm) are of vital importance. But they are not the whole foundation of the Algorithmic Wars.
The enhanced ability of computers to predict behaviour goes far deeper than serving up adverts, or suggesting news stories. There is a predictive power of conversations on social channels to indicate future action, e.g. protest turning into violence. Or the ability for Organisations to scan social channels to predict who of their staff is most likely stealing, or rousing dissent. Our ability to ‘listen in’, to ‘predict’, to sense the location of tipping points, all of this is evolving.
Unless we smash the looms, unless we choose to reject the many benefits of these new technologies, we are just at the start of a long, and evolutionary process.
There will be many mistakes along the way, and some people will make a very great deal of money, or achieve significant influence and power, by exploiting the new dynamics faster than we can regulate, or even notice what it is that is happening. But none of this makes it all bad or wrong.
As ever, our challenge is this, and it’s a challenge we must face up to in the middle of this war: technology will take us into places that we are ill equipped to deal with. But our ability to deal with it cannot be framed in the old understanding of knowledge, decision making, and power.
It’s a new type of challenge that is faced in a new kind of space.
And it will require new types of thinking to ensure that, on balance, the change takes us into a new type of space that we can comfortably inhabit. Primary interpretations of the current swathes of change according to well known and well understood frameworks may be dangerous: it may comfort us to think of small groups of elite enemy agents undermining our democracy, but this is but one facet of change.
The real outcome of the Algorithmic Wars may be decided through schism and conquest, but most likely will be an outcome of optimisation and greed: the ways we engage with knowledge, the ways we shop, connect, think, act, all influenced by myriad underlying algorithms. An unknowably complex series of filters and moderators of individual action: a radically complex set of predictive engines, and all continuing to learn, to evolve, in a tumbling wheel of change.
Perhaps our greatest challenge is to find ways to narrate, and understand the sheer scope of the challenge, and to articulate what it means for us as individuals, for our Organisations (which have so much to gain, and so much to lose) and for wider society as a whole.