Two very different stories have been prominent in my twitter timeline over the last few days. Despite their differences they illuminate each other and help shape answers to the perennially vexed question of what constitutes privacy in an online world. One is about our licence to analyse people’s behaviour, however benign the intention might be. The other is about what it’s like to walk down the street. They have more in common than it might first appear.

Stalking on twitter?

The first is Samaritan Radar.¹ If you’ve missed it, it’s a new service which allows a twitter user to ask the Samaritans to monitor the tweets of people they follow (who may or may not be personally known to them) and to email alerts if any of those people show signs that they ‘may need your support’. To put it mildly, that has been immediately contentious. The subject of this analysis is not told that it is happening and has no idea that the Samaritans may be making judgements about their mental state, still less broadcasting them to people who may be complete strangers. I am not going to go into the reasons here why that might be a thoroughly bad idea – Adrian Short has three great posts which explain why both this particular service and the general approach it embodies are highly problematic. Meanwhile, the Samaritans have published an update, supposedly responding to the debate, but not in a way which really addresses the criticisms which have been made.²

But is all that missing the point about the openness of twitter? Radar can only work with twitter accounts which are open, anybody can find and read anybody else’s tweets, so they are the very definition of public.  And if people speak in public, the argument goes, they can hardly complain if they are overheard. I think that view is wrong, fundamentally because it assumes that technical capability does (or should) map to social acceptability. There are three big reasons why that is not so.

Twitter isn’t simple

There is a human urge to put things in pigeon holes, to impose neatness and regularity on the world. But that can lead to confused and misleading results if applied to something which isn’t as tidy as that. I can’t improve on the description of twitter I first used several years ago:

For twitter in particular, there is a very strange collision of contexts. It is like being in the pub with some friends, being at speakers’ corner shouting at (and being heckled by) random passers by, being on the Today programme, being on Big Brother, and throwing a message in a bottle out to sea – all at once.

Twitter is different for different people, and different for the same people at different times. Radar assumes the friends in the pub case (though even that is not without its problems) and chooses to ignore the rest.

Being heckled by random passers by has become a nastier pastime since I used those fairly innocent words. Misogyny is rampant, threats have been all too real – which makes it remarkable that part of Samaritans’ response to the concern has been to assume all that away:

The aim of the app is to look for potentially worrying tweets from people talking about their problems with the hope that their followers will respond to their Tweets – which are already public – and which otherwise may be missed. Those who sign up to the app don’t necessarily need to act on any of the alerts they receive, in the same way that people may not respond to a comment made in the physical world. However, we strongly believe people who have signed up to Samaritans Radar do truly want to be able to help their friends who may be struggling to cope. [Emphasis added]

That’s a fairly breathtaking belief, if for no other reason than that the service works by monitoring all the people you follow, so even with the most benign motivation it isn’t fine grained enough to avoid providing updates about complete strangers whom you may be unwilling or unable to help in any way – perhaps, as in the example below, because they have already been dead for some time.

Of course my first #SamaritansRadar alert had to be this guy pic.twitter.com/GePXlS78PA

— Stef Lewandowski (@stef) November 3, 2014

That takes us neatly to the second reason.

Privacy is not binary

Technically, anyone can follow anybody on twitter, all tweets are public (except when they are not) and it is all part of the panopticon. Privacy is simple: there isn’t any. Except that that isn’t the perceived experience at all. I may have no means of preventing somebody obsessively analysing my online presence, but that doesn’t make it either socially expected or socially accepted (‘but it’s legal’ isn’t terribly relevant here: there are all too many ways in which behaviour can be entirely legal while simultaneously obnoxious and objectionable). The experience of most twitter users most of the time is that their tweets are read (if at all) by a small number of people.³

Technology has a bias to the binary – your tweets are public or private, links are alive or dead, your location is visible or invisible. But  real world behaviours and expectations aren’t like that, for reasons David Weinberger captured long ago in his concept of ‘leeway’:

Leeway is the only way we manage to live together: We ignore what isn’t our business. We cut one another some slack. We forgive one another when we transgress.

Radar is not negotiating this minefield of tangled assumptions, slowly adapting to changing circumstances.  It is just ignoring it. Bizarrely, their privacy page concerns itself purely with the person doing the monitoring and not at all with the person being monitored.

Of course there is the counter-argument that anything which can be done technically, will be done, summed up in Scott McNealy’s famous line from 15 years ago:

You have zero privacy anyway. Get over it.

But whatever level of privacy we have, it clearly isn’t zero, in part because we create social limits and expectations in relation to what is practically possible. Privacy has levels, not states, and it is reasonable – even necessary – to ask what level is appropriate for a service such as Radar.

Many small things make a big thing

Aggregation transforms. Individual tweets may be willingly public, but it does not follow that the analysis of a body of tweets inherits the permission belonging to each tweet individually. In the UK at least, that’s more than just a social norm (though it’s being a social norm is again separate from its legality).  As Zoë Kirk-Robinson points out, it isn’t reading or even reproducing or retransmitting tweets which is the problem, it is aggregation, analysis and dissemination which put them into difficult territory:

It’s true that tweets are public information but the Samaritans are not just passing on your tweets, they are processing the “emotional content” of the tweets and determining your mental state as a result.

It is this information – their data-mined analysis of your emotional state – that is being transmitted to a third party, not your tweets.

Harassment in New York

That brings us to the second story. A woman is filmed walking along the street in New York. Ten hours of walking is boiled down to a two minute video, in which she doesn’t say anything or do anything except walk. That video was published less than a week ago, since when it has been watched more than 30 million times. That’s because it distils what feels like relentless harassment, prompted by nothing more than a woman walking by herself.⁴

None of the incidents is overtly aggressive (though at least one is decidedly spooky), and a world in which any one of those incidents happened might not be a very terrible one. But a world in which they all happen, predictably and relentlessly, is one nobody should be required to accept. Nobody in the video does anything which appears to be – or should be – criminal, but cumulatively they amount to harassment (and if all those things were done by one person, their behaviour probably would be illegal).

But in the more familiar environment of the street, we can see versions of the same three factors which confuse things in twitter. The parallels aren’t exact and shouldn’t be strained too far, but do bring out some of the same tensions in a more familiar and far longer established context. In our offline lives, we don’t tend to think that what is possible is thereby right, or that what isn’t illegal is unexceptionable.

The street isn’t simple

Streets are for walking along. Streets are places to  have conversations. Streets are places to have demonstrations.  Streets can be places for very private exchanges. Streets can be places for addressing large audiences. That’s so obvious that most people don’t stop to think about it, and have little difficulty in understanding that there isn’t a single right way of interpreting what’s going on – though there are some which are just plain wrong.⁵ Responding to what goes on on the street based on assumptions about common purpose and motivation is unlikely to be effective.

Privacy is not binary

If I am walking along having a quiet conversation, you are not entitled to eavesdrop. You might overhear, but going out of your way to listen or follow – and particularly going out of your way to listen or follow in a way which those in the conversation are unlikely to notice – would be widely seen as socially unacceptable. Increasingly, technology makes it possible to breach that social norm – directional microphones, drone surveillance, phone tracking and more are provide the capability of eavesdropping, but don’t (at least in the short term) change our view of its acceptability. Charlotte Walker has written a post using the analogy of a café to make a similar point far more powerfully.

Many small things make a big thing

People should be able to walk around without getting harassed by anybody. One small thing is already one too many.  But aggregation transforms. The aggregated impact is very different from each of the individual impact.

And so

So where, if anywhere, does that get us?

As far as the Samaritans are concerned, I am pretty sure that Radar just gets it wrong. Having technology which enables you to do something doesn’t make it smart to do it, and this particular idea seems remarkably poorly thought through.

But I am more interested in two broader thoughts.

The first is that expectations are very inconsistent. Part of why people feel cross is that they expect better of an organisation which not only has a social purpose, but a social purpose based on empathy. Meanwhile, less worthy organisations are amassing every kind of information they can about every one of us, for less laudable and less obvious purposes. As Bruce Schneier puts it

Surveillance is the business model of the internet.

The second is that the social norms of social media are still very fluid. Mapping real world social relationships and patterns of behaviour onto the virtual world sometimes works, but very often doesn’t. That’s not helped by the appropriation of words from one context to use in another – what twitter and facebook mean by the word ‘friend’ isn’t what I mean by it (and that gap may be part of the confusion behind Radar).

Those two thoughts are obviously closely related. Whether we choose to address them as social questions or technical questions will make a big difference to the answers.