Thursday, 30 November 2017

Distributed social networking in 2017: a review

Abstract

In just the past ten years or so, use of social networking sites has grown to the stage where the market leaders can claim billions of active users. In some areas, social networks go far beyond simply connecting friends, but are users' primary source of news. Some social networks can even drive the traditional news, having a measurable effect on emerging events.

Suffice it to say, then, that social networking is perceived as being as essential to many users as access to a TV, radio and landline phone would have been to our parents' generation. Yet a growing number of influential commentators and individuals are belatedly starting to realise that there is a data privacy trade-off inherent in using social network sites. Some networks (Google, Facebook, Twitter) are so ubiquitous that they are able to track users' activity on sites across the web, even when the user is not not logged in, and sell this browsing data to advertisers. We are encouraged to share more and more personal information with these sites, on the basis that this allows our friends to connect with us better and for more targeted suggestions of other content (i.e. adverts). This is known as "surveillance capitalism".

Beyond the privacy issue, there are concerns that old, potentially embarrassing data may resurface in the future and can never be truly deleted. There are concerns that the algorithmic selection of what to show in a user's feed means that people's perception of world events is being warped (called a "filter bubble"). There are concerns about centralising data in a country with extensive state powers to access that data covertly. And there are concerns that data is being used in inappropriate ways; for example, personal photos appearing as an integral part of an advert, to falsely imply a personal endorsement.

What if there a way of keeping the features of a social network that make it useful to people - for example, the instant ability to make all your friends jealous of your holiday - but bake in privacy and control, to eliminate the types of undesirable behaviour listed above? Actually, we can already do this. The method is called distributed social networking, also known as decentralised or federated social networking. It's not well-known outside of a few niche circles, but it exists, under active usage and ongoing development.

This review looks at two of the most mature distributed networks in an attempt to determine whether they will meet the needs of an average user.

Background

Anyone who has read any mainstream media recently will know social networking is responsible for making millennials stupid [citation needed], and is only ever used for broadcasting pictures of our breakfasts and / or our own legs in front of a tropical sunset #SoBlessed #AnotherPinaColadaPlease #PleaseLikeThisSoICanFeelValidated. A growing number of people are starting to recognise the uncomfortable privacy trade-off inherent in social networking as well: that we freely give up our personal information in exchange for access to a proprietary, closed communications system that, by design, spies on our most intimate moments, thoughts and preferences. They do this in order to sell us things and even to influence our own core beliefs.

I downloaded my personal Facebook archive. I'd consider myself a light user, yet the archive exceeded 40MB and more than a thousand files. That's not including all the analytics data, the proprietary decision-making processes that enable Facebook to determine what kind of advertisers might be interested in reaching me. The archive contains all of my contacts (current and deleted), all of the events I've attended or ignored, all of the private message threads that I've participated in over more than a decade. It contains every ill-judged "joke" and comment, every embarrassing photo in which I have been tagged. It contains adverts I've recently clicked on (usually by accident).

I don't mean to pick on Facebook, whose enormous success is down to many factors, including having engineered the most advanced software platform of any social network in the world. When my friends and I use Facebook, it is because it fulfils a useful function for us. We generally ignore the downsides, much like a meat-eater can put to the back of his mind the harsh realities of an abattoir. However, it is telling that the Wikipedia article entitled Criticism of Facebook runs to some 20,000 words in length, excluding its references.

It doesn't have to be that way. It is possible to have all the features we understand as being necessary to a social network, but free. That's free as in "beer", free as in "speech", free as in "you can leave whenever you want", and free as in "won't insinuate itself into your conversations, shouting at you to buy stuff."

There are a number of open-source software projects that do social networking in a way familiar to any user of Facebook or Twitter. Some are thin veneers over other systems, offering additional functionality or some privacy protection but communicating over the underlying closed platform. Others are complete clones or imitations: Mastodon, for example, is basically the same as Twitter, except that the leader of the free world isn't doing diplomacy via Mastodon. And Mastodon has one huge privacy-friendly feature that Twitter doesn't have, and which we will be discussing in depth in this article: federation.

How is this possible? Surely Twitter's unique selling point is that it is, in fact, unique? Well, yes, kind of. If you are looking to interact with celebrities (or, indeed, just your own friends), then clearly you need those celebrities (or your friends) to be on the same platform, or at least, on one that is compatible. In that sense, Twitter is unique.

But when the tabloid press regularly blame social networking for all the world's ills, the key mistake that they make is to view all social media as some kind of ultra-advanced technology, only really understood by the kids. It isn't. Use cases vary, of course, but in principle a social network platform is simply a communications tool that allows ordinary users to upload "content" - words, pictures, videos - and for other users to respond in kind. Most newspaper websites' comments sections can do that. Simple blog software like WordPress can do that. But these other sites haven't scaled to build vast communities out of their contributors. The thing that really makes the major social media platforms work is that they are monopolies. And like any monopoly, especially one with shareholders, they can be abusive.

(It's worth considering how they became monopolies, too, and think of all the other potential monopolies we might have had instead. Friends Reunited, Bebo and MySpace all had a far richer feature set than Twitter did at launch. Famously, most of Twitter's most well-known features, such as the convention of hashtags, were developed spontaneously by users as a workaround to the platform's own shortcomings.)

There are alternative approaches to subverting this model. One way would be to develop software modelled on the features of, say, Facebook, but without centralising all the data in a way that makes it ripe for exploitation. Another way would be to build tools that minimally interact with the platform but which exercise additional controls over the top.

I've already hinted at the solution. WordPress proves that the publication technology not only exists (for free), but can be packaged up for novice users to install on their own websites. It's a small step from there to building a trusted network of virtual publications, whose users can interact with one another. This is software that doesn't just connect publishing platforms together, but manages contacts, interactions and content permissions in a way that preserves the user's preferences and privacy, while also providing near real-time updates to contacts on other sites and networks.

In this model of federated content platforms, users can choose to host their own servers if they feel strongly about privacy and have some technical skills, or they can choose to sign up to a server that somebody else is hosting already. For the purposes of this review, I looked at two of the biggest projects offering this model: Diaspora and Friendica. These two networks can be federated together, so somebody on Diaspora can be friends with somebody on Friendica. This became very handy for testing. They aren't exactly identical, however, so this review will attempt to identify the strengths and weaknesses of each. Also, both projects are under active development, so features will change and improve over time.

How it works

Both networks have a similar sign-up procedure. As a user, you can go to a list of nodes on the network (Diaspora calls them "pods"), select one you like the look of, and sign up. You might choose it because it's based in the same country or because it's been recommended to you or because it has useful plugins or because the graphics are pretty. But you definitely don't have to choose it on the basis that your friends are on the same pod already: all of the nodes communicate with one another.

Or, if none of the nodes take your fancy, you can build your own. Download the software, install it, configure it. You, and you alone, are then responsible for the node's security and all of your own content. You can choose whether to let other people use your node or you can keep it for yourself.

For testing purposes, I created an account on a Diaspora node based in Norway, and an account on a Friendica node based in Germany. I quickly discovered that having two accounts with the same real person name made for headaches when pinging test messages back and forth, so I doubled-down on this mistake by creating two more accounts with an identical pseudonym. My fictional friend Felicity and I, and my other fictional friend Felicity, and the other I, all became friends with one another and then spent a couple of weeks posting content back and forth between our two Diaspora and two Friendica instances. Yes, I spent a fortnight talking to myself for the purpose of this review. It is possible that I need more real-world friends.

Usability

Both Diaspora and Friendica use familiar timeline-based activity streams. Both allow more stream management than Facebook does. For example, depending on how you have classified your contacts, you could view just your family's posts, or just your colleagues' posts; or all public material on a particular topic; sorted by most recent or by most relevant; and so on. It's much more flexible than Facebook's curated approach and, of course, you definitely won't have third-party adverts appearing in the stream, masquerading as content.

This works the other way as well - whenever you post, you can choose which groups or individuals should see your posted content. You can even have multiple personal profiles reflecting different elements of your life, and you can configure your account to work either symmetrically (for example, friends who mutually share with you, as in Facebook) or asymmetrically (for example, for fans to "follow" your work, without you following them back, as in Twitter).

Friendica offers some more functionality than Diaspora, but it's sometimes a bit clunky. Here are some features that Friendica has that Diaspora does not:

  • Calendar and shared events
  • Photo albums
  • Edit existing posts
  • Threaded conversations under posts
  • Subscribe to RSS feeds
  • Move accounts between nodes (this feature is described as "experimental", and did not work very well when I tried it towards the end of testing; it did not deactivate the old account, so that I ended up with two functional instances of the same account)

Features of Diaspora that Friendica does not have:

  • Private messaging to multiple recipients at once (Friendica's UI doesn't support this, but the protocol does - so perhaps it is coming soon)
  • Built-in instant message (XMPP chat) support, integrated to Diaspora contacts

Both networks allow you to plug in or connect to other platforms. For example, you can cross-post from Friendica to Facebook or Twitter. In the other direction, I tried a WordPress plugin that posts blog entries straight to Diaspora, and it worked fine. The import and export services available on any given pod or node will differ according to the node admin's preferences and patience.

Look and feel

In my view, Diaspora has a more professional appearance, but also a more austere one. Friendica's interface is a bit friendlier but seems unfinished in places.

The software versions of the two Diaspora instances I tested were different, but the overall look and feel was very similar. There are some basic customisation options for colours and layouts.

Conversely, the two Friendica instances I tested had quite significantly different default themes, despite being on the same version of the software. Friendica also allows customisation of colours according to pre-defined templates or even custom colours for basic page elements.

As part of testing, I installed Android clients for both networks. Both of the Friendica ones I tried were simply terrible - unfinished, unreliable and ugly. The main Diaspora app is a thin veneer over the website. It works well enough, except that it thoughtlessly draws over the website's perfectly functional notifications area with a tool bar in which notifications don't work. In other words, the experience is better on the websites of both Diaspora and Friendica than in their apps. Both websites have mobile modes and respond well to smaller screens.

Security

All your contacts can be assigned to a group (Diaspora calls them "aspects") such as Friends, Family, Work. You can choose which posts are visible by which of your contacts or groups, or you can post publicly.

I found the default permissions on Friendica to be a bit misleading at first - although some of this was simply down to my lack of familiarity. Once I'd realised my mistake, I also found that retrospectively changing the permissions caused some strange side effects, including content disappearing for some of my established contacts. Additionally, some activity that I erroneously posted publicly remained visible after I deleted the associated content.

Your profile information and comments will always be available publicly if you respond to a public post.

Reliability

In the course of my testing, I took pages of notes. It would not be particularly interesting to relate all of the bugs I found. Both platforms had their share of quirky behaviour, but they generally worked as intended (once operator error was eliminated). Friendica had a couple of incidents in which post edits weren't saved, which is partly what leads me to believe Diaspora is marginally more robust.

Some functionality did not appear to work at all, but this could be a problem with the individual instances. For example, Diaspora includes a chat widget but I could not get this to work, which I suspect is a problem with my particular pod's XMPP configuration.

Federation

Contacts on Diaspora and Friendica, along with certain other platforms, can communicate with one another natively, although within the feature constraints of the respective platforms. For example, Diaspora users won't see threads in comments under posts, and won't see shared calendar events.

Some of the differences between the platforms are much more frustrating and there are no obvious workarounds. For example, a photo published on Friendica as anything other than fully public will not appear on Diaspora, even if embedded as part of a larger post that is visible to contacts. The Friendica and Diaspora permissions models are not compatible. In contrast, Diaspora makes photos available via a secret URL; anyone with the URL can see the photo. Therefore, Friendica users can see Diaspora photos.

Other than that issue, generally, Diaspora seems better at receiving and caching data from other nodes. This makes it more resilient to nodes becoming temporarily unavailable. By contrast, Friendica tends to assume that content on other nodes will always be available; if a node becomes unavailable for some reason, then profile and content on that data will temporarily disappear from feeds on other nodes.

There are some particular quirks around subscriptions to public posts. Diaspora users subscribe to topics of interest using a hashtag and will see public posts with those tags in their timeline. But the exact posts seen vary by pod. For example, both Felicity and I subscribed to #italy and our timelines were similar, but not identical. I understand this to be a peculiarity of the way that Diaspora federates public posts that aren't from users being directly followed.

Diaspora also suffers from moderate levels of pseudo-spam in the form of unwanted public posts. It's easy to block a user, but I wonder whether this will escalate in future. Spam posts may be from well-meaning but prolific users who have tagged their content badly; or it may be that I am being too general in subscribing to a generic term like #technology. There are also "bot" accounts on Diaspora that take content from third-party sites, attempt to classify it, and then re-publish it, with mixed results.

Self-hosting

In signing up to a node hosted by someone else, you are placing your trust in them. They are probably hosting the node for their own fun and education. They probably don't have a complaints department when things go wrong. There is a chance that they will lose your data or simply shut down without warning.

For the privacy-conscious user, the ultimate goal of joining a federated social network must be to take personal control of a node and all its content. I did not test this, but did read through the instructions for installing both. I am personally comfortable working on a LAMP stack, as used by Friendica; but less familiar with the Ruby framework on which Diaspora is built. The Friendica instructions are written in plain English and targeted at the level of someone with familiarity with configuring WordPress, Drupal or similar on a shared hosting package. Overall, the Diaspora instructions felt rather more complex and the system pre-requisites greater. I do not believe that Diaspora could be installed successfully on a shared hosting account; a VPS would be the minimum requirement.

It is not clear to me how much system resources (hard disk storage and bandwidth) each network would consume. It is also not entirely clear to me how the administrator of a node would set about moderating content, ensuring legal compliance etc. For these two reasons, were I to set up a personal node, I would not allow the public to sign up. I appreciate that this attitude is not strictly in the collaborative spirit of these networks, especially as I have taken advantage of four nodes whose admins have been willing to do exactly that.

The effect of other users on the same node can be unexpectedly far-reaching. The administrator of one of the nodes on which I have an account posted a public message to the effect that the Diaspora-to-Twitter connector would no longer function. One of the pod's users had breached Twitter's Ts&Cs and now all users from that pod were blocked. I think this is a clear over-reaction from Twitter, who would have had the individual user's credentials and could have blocked just that user, but it demonstrates a level of brittleness in the component model.

Volume, reach and retention

It is extremely hard to quantify how many active users are on each network. There is a site that attempts to collate these statistics, called the-federation.info. Its headline figures are deeply disappointing. Friendica shows 403 active users. That's not a typo; there are no missing thousands or millions here. Diaspora has more at around 16,000, of whom more than a quarter are on a single pod.

The problem with collating this data is that it relies on statistics collected across a federated network. Not all nodes provide accurate data, or any data at all. I think we can safely say that the-federation.info is underestimating the number of active users. However, even at our most optimistic, we are still many orders of magnitude away from the reach of the centralised leaders.

Worse, the number of active users vs total registered users shows that the vast majority of people who sign up do not stick with it long-term. If they cannot retain early adopters - the most privacy conscious people, or those who have been nudged into joining via their communities of interest - then it seems unlikely that the networks will grow in the long term. (Some Friendica servers automatically delete non-active users, so the numbers of active and inactive users tend to track one another.)

Yet when signing up for a Diaspora account for Felicity, it actually took a number of attempts to find a valid username that had not already been taken. Felicity is not exactly a common name in Germany, where the pod is located. This is consistent with large numbers of no-longer active users; perhaps those who, like me, signed up out of curiosity but without a long-term intention to stay.

Friendica has a centralised directory of people who have chosen to opt-in to sharing their details. Again, it's a depressingly small volume, measured in the hundreds. Of these, the largest population is from Germany. Unsurprisingly, a large fraction of those in the directory have #linux in their profile. In fact, Linux is a good analogue for Friendica: it's mature enough to be functional, free (beer / speech), yet fails to appeal to the mass population. Linux has been consistently touted as being ready for widespread adoption for years. Similarly, Diaspora and Friendica have both been touted in the (technology) press as being good alternatives to centralised social media since at least 2012, but have apparently not made the impact they deserve.

Use cases

I don't personally use social media for consuming world news. I use it to stay in touch with friends, especially those with whom I used to be close, but now live far away or are busy with families and work. My main criterion for connecting with someone on Facebook is: if I met up with this person tonight, would I offer to buy them a beer? I don't, personally, expect to reach large audiences of strangers through social media, largely because I don't want to end up on either side of a political flame war. I'm certainly not going to win thousands of followers on the strength of this blog or my so-called poetry.

Sometimes I want to share pictures of my family in a way which preserves our privacy. In the past, I have done this through a personal website with password protection. This still seems to be the favoured approach for my technology-literate friends at the time of a major family event such as a wedding. In this case, you don't need to authenticate particular users, merely restrict access to those who know the password. Diaspora and Friendica would fail in this use case, because they would require each contact to sign up to a service that they don't really want. Also, due to the way federation works, they would have to become a mutual contact before publishing any useful information; it is not possible to view information published before the sharing relationship began.

In the workplace, we use several different social / content platforms. Each is a walled garden. SharePoint for corporate data; Confluence for engineering data; both Skype and Slack for instant messaging. Lacking the more persistent content management aspects, Diaspora and Friendica would fail this use case, too (although the Friendica-related Hubzilla project might be worth a look).

There may be niche areas where the federated approach helps to protect the identities of activists and whistleblowers, putting them out of reach of the (mainly US) legal system. The ugly flipside of this anonymity is that they would also offer a haven for criminal activity. And the distributed nature of the network doesn't necessarily help here either: it would be far easier for a legal authority to block a Friendica node on the basis of alleged illegal activity, than it would be for it to take down the whole of Twitter. I cannot test this, of course, but I suspect Diaspora's sharing model to be slightly more robust against this eventuality than Friendica's.

After much thought, then, I have yet to come up with a compelling use case that not just caters to ordinary users, but would attract them to switch away from the incumbents in large numbers. Indeed, the only way I can see either network really turning their meagre toehold into the critical mass that they deserve is if a large, federated organisation suddenly decided to endorse the network. For example, we might imagine a scenario in which the National Union of Students mandates each member union builds and maintains a Friendica node. Students could sign up to their local node to stay in touch with friends at other universities. When they leave the university, they could migrate their profile to an alumni node or their own preferred server. This hypothetical growth would nicely mirror the original growth of Facebook, which started off as an invitation-only network in US universities.

If you regularly cross-post to several social networks, and consume RSS feeds, then you could consider signing up for an account on one of the platforms to use as your main publishing home. Perhaps, in the future, more people will migrate and you might make some new friends on the platform.

For everybody else, and with some regret, I must conclude that Friendica and Diaspora are just not ready for you to use in earnest. Not because the software is immature or non-functional, but simply because your friends aren't ready to follow you there.

Credits

Many thanks to the owners / admins of the open sign-up servers that I used during testing, who are paying for my experiments in terms of bandwidth and storage costs. Those servers were: EgeLand (David, Diaspora, monkey avatar); JoinDiaspora.com (Felicity, Diaspora, Fio avatar); friendica.me (David, Friendica, Buddha statue avatar); and nerdica.net (Felicity, Friendica, Ronja avatar).

Further reading: perspectives on using Friendica in conjunction with established networks

Linux Magazine, "Developing for a Post-Facebook World" by Bruce Byfield. Comments from the lead developer of Friendica: "I currently interact daily with friends on Facebook, Twitter, Diaspora*, Identi.ca, and Friendica - all from within Friendica. I also have friends in my stream who only have email addresses and RSS feeds ... It shouldn't matter if your friends use Facebook or Google+ or Friendica or Diaspora or anything else. They're all just pieces of software you use to access your social communications. We want to break down the walled gardens and show them for what they are: corporate walls that were built for business goals and actually prevent you from communicating with friends, unless you become a member of every different service."

Clear Linen Tea blog, "Friendica" by Sonata Green. Relatively recent (2016) blog post argues that Friendica offers a superior and functional way of connecting social networks together, as well as outlining some of the objections raised above. "Friendica can, like Diaspora*, post to traditional social networks. Unlike Diaspora*, though, Friendica can read from them as well. This two-way connection means that, using Friendica, I can engage in conversations with people on Twitter and tumblr and Diaspora*, all through a single unified interface on a single site that I control. Furthermore, these different networks aren't just collated - they're integrated. Because my Friendica is connected to both my Twitter and my tumblr, this means that my Twitter and my tumblr are - through Friendica - connected to each other."

Sunday, 1 October 2017

100-word fiction: Time to leave

Hi, Mum. Sorry we haven’t been to visit. How have you been?

Did you see the news this morning? Yes, it looks really bad this time.

I’m going to take the children away from here. There’s a group leaving any minute. I have to go straight away – they are waiting for me outside.

I thought of staying longer, but it’s not a good idea to wait. I can’t promise that we will be safe where we’re going, but I’m certain we aren’t safe here.

They are calling for me now. I have to go.

I love you, Mummy.

100-word fiction: Look who's back!

He's just there, sitting gormlessly at the bare table. He looks at me, eyes empty, gulps and says nothing.

"Look who’s back!" she breezes, fussing around to make him comfortable. A teacup and saucer clatter down. I jump. He doesn't.

He hasn't aged. Ten years and he looks precisely the same.

My fists clench. Offended that he should reappear like this; sickened that he left in the first place.

He is not real. He is an imposter. He is a ghost. I am angry and I am scared.

Her face warns that my urgent questions must wait.

Wednesday, 8 March 2017

Why I support the expansion of grammar schools

I try to ensure that my blog is generally light-hearted, entertaining and apolitical. I strive to avoid causing offence. So it's disconcerting to find myself on the deeply unpopular "wrong" side of the socially-progressive consensus, arguing against organisations such as Reform and the Institute for Fiscal Studies. It feels very much like the first time as an adult that somebody told me, to my face, that they didn't like me.

Yet here I am, arguing for the expansion of grammar schools. A policy of a government that I didn't vote for, strongly advocated by a Prime Minister that I don't much care for.

And yes, I have been goaded into setting out my views because of this kind of popular argument:

I would say to @twlldun (in rather more than 140 characters):

  1. Absolutely, I acknowledge that my own singular experience is only an anecdote. I still have the right to talk about it.
  2. Your comment is an unacceptable generalisation - just as flawed as the poor grasp of statistics that you ridicule.
  3. You can't extrapolate statistics about a hypothetical future education system from the system that existed in the 1950s.
  4. Who says that you're asking the right questions of the right people? You could pose a question to a population of grammar school alumni - say, "Was your experience of grammar school a good one?" - and still get meaningful statistical answers. (I don't know whether anybody has ever asked such a question, but it seems likely that that is the implicit question being answered in many such anecdotes.)
  5. Even within a grammar school education, not everybody is going to end up being great at statistics.

Here's my anecdote, then. I wouldn't be in the position I am today were it not for a grammar education. I wouldn't have been the first in my family to go to university, closely followed by my little sister. My wife and her brother would say the same. As would many of our friends. My parents did not go to university; but they both entered professions as a result of being the most academically gifted in their families and, consequently, attending grammar schools.

Did this contribute to a split along social lines? As it happens, us going to grammar school did not somehow magically destroy our friendships with those that didn't. We stayed in touch with those people we cared about - those with whom we had common interests. Of those that I regret not staying in touch with, a fair number went to other grammars, especially one of the two local girls' grammars.

That's not to say I think my grammar education was perfect. I think a single-sex education, while possibly conducive to academic concentration, had a strongly detrimental social effect on me and many of my peers. More fundamentally, some of my classmates have subsequently made a compelling case that they became complacent in education: that the school failed to enable them to reach their full potential; that it did not add as much value as it could have. There is some evidence for this: at the time of my attendance, my school was found to be the "best value for money" state school in the country. This means that it spent less for every high GCSE or A-Level grade than any other. You could take this to mean that the school's high achievement was despite, not because of, high educational standards.

Still, I think I would have rapidly foundered in a comp. Bookish, nerdy, resolutely uninterested in sport, and introverted: I think I would have been bullied, that I would have been easily bored in lessons, and that I would eventually have retreated into myself. I'm not sure about any of these hypotheticals, but I believe them. Around the ages of 11 or 12, I was twice attacked by older kids from local comps because of the uniform I wore. I quickly came to see that my school was, relatively speaking, a haven of bright pupils whose most significant common factor was that they all had a healthy respect for school and for learning. I agree that they weren't an especially diverse bunch, but then, I'm not sure that their backgrounds differed significantly from those friends I had at primary school.

I am aware of the main arguments against selective education. It doesn't increase social mobility as much as it should. Only a tiny elite get to attend. It leaves behind bright pupils whose potential has been overlooked. Overall attainment decreases. It places too much emphasis on the outcome of a single test, when we all develop at a different rate. Richer students are more likely to succeed because they can afford private tuition. I believe that all of these problems can be overcome with a well-designed grammar system. Indeed, the Government appears to be doing exactly that in its consultations.

Social mobility can be addressed either through quotas (though these are fraught with difficulties); or, as the Government proposes, by having the school actively involved in under-privileged feeder schools. The argument that an insufficient proportion of students are eligible to attend a grammar school can be addressed by providing more places, i.e. by expanding grammar schools. And there's no particular reason why there has to be only a single point of entry. My school accepted entrants at 11, 13, 16 and at other points in between.

Overall attainment in a selective educational area decreases only because the quality of other schools is so low. That is not inherent in a selective education system, although it might well have been in the past. Idealistically, we should aim to improve all schools - catering to all needs. There is no need for us to choose between grammar schools on the one hand, and raising standards at comprehensives on the other. There are even proposed mechanisms for ensuring that this happens within a finite budget, such as mandating that the grammar is part of a multi-school trust. And we can be smarter about how we measure success, too, recognising that a student's lower academic ability does not equate to failure.

The problem of a single entrance exam, and the possibility of richer students being coached to pass, can both be overcome by use of different entrance criteria. In fact, I believe that the problem has already been solved once before. At the time of my own 11-plus exam, teachers told my parents that the procedure was roughly this: the school marked coursework for students over a period of years, identifying the most able. The exam was then used to benchmark schools against one another. Finally, the highest-ranked pupils from each school were selected. My parents were told - twice, a few years apart - that both my sister and I could have missed the exam entirely and still be certain of a place at grammar school. Coaching would have made no difference either way. I am very surprised that this approach is not documented anywhere. I can't believe that the teachers lied to my parents, especially when the mechanism that they described makes more sense than the current one-shot test.

Why do I support the expansion of grammar schools? Because I believe that it would be unconscionable for me to argue against a system from which I have personally benefited so substantially. Just as I benefited from a free university education, and hence disagree with the principle of exorbitant tuition fees; and just as I have benefited from the NHS, and therefore support that organisation (despite its obvious flaws). Let me repeat: I understand that my grammar education was an opportunity and a privilege, and notwithstanding all arguments to the contrary, it would be unethical for me to want to deny the same privilege to others.

It is widely reported that the current younger generation is the first in decades to be worse off than their parents. The generation now approaching retirement have systematically pulled up the drawbridges behind them, cutting off younger people from the decades of social progress from which they prospered. I think we owe it to the younger generation to reverse this and open up opportunities for them; to give gifted children from less well-off backgrounds a genuine chance to succeed. Equality of opportunity does not mean that we have to place the most gifted students in the same classroom as the less able, and those who simply don't want to learn; but that we strive to design an education system that benefits all according to their unique needs.

The selective system is not without its flaws, but it is better to fix those flaws than to resign ourselves to a mediocre average.

Monday, 20 February 2017

366 giorni della lingua italiana

For the past year, I have been learning Italian, practising a few words every single day, using an app called Duolingo. (In fact I started more than a full year ago, but the app allows for me to take short breaks from time to time.)

Apart from 1,500 words of vocabulary, here are a few things I've learned about the process and about myself over this period.

It's much easier to learn a language when you're motivated - even if the rationale is spurious

Why did I choose Italian? No greater reason than I love visiting Italy.

I don't think learning Italian is likely to make a difference to my career prospects. I recognise that there are many other languages that I could have picked that would be more useful in a globalised economy. Yet Italian is the fourth most-studied language in the world.

Surely most students of Italian have chosen it with their hearts, rather than their heads. Like many of them, I have been seduced by the place. I have an ill thought-through fantasy that I will take early retirement to buy a little farmhouse in Umbria, where chickens roam freely among the vines and where I can press my own olives.

At this point, I was going add a little "top ten" list to explain why I love Italy so much, but I can't realistically limit myself to ten items.

But it's worked. I've stayed motivated. I've graduated from an audio course, to a phrase-a-day calendar, to daily lessons.

My wife, by contrast, did well at first but didn't stick with Italian. But she's now found a different language that she's interested to learn, and has her own motivations for doing so - and has notched up a very good run already.

Gamification works

The Duolingo app is actually kind of basic - a small selection of different types of exercise. Its success is down to two things: first, a thoughtfully graduated selection of vocabulary along with a successful algorithm for selecting words to be practised; and second, gamification.

Gamification means that the app treats learning like a kind of game, with rewards for completing particular tasks. The gamification elements of Duolingo include Experience Points, Levels, Badges, and a virtual currency. The virtual currency, called Lingots, is, of course, totally worthless in the real world, as are the virtual things that it buys.

For me, the most important gamification element is the "streak": a simple count of the number of consecutive days that I have reached my daily goal. That now stands at exactly one year. I am disproportionately proud of this fact and I would genuinely be crushed if it were to reset for some reason. For me, that's all the motivation I need to make sure that I do a small amount of practice every day.

And the instruction method itself does, indeed, work. I am certain that I am retaining more information through brief daily practice sessions than I would do if I tried to study for a full hour, once a week.

Different media helps, but learning a language through audio course alone is very hard

The Duolingo experience on desktop is quite different to that provided by the app. Using both together is ideal. The mixture of reading, writing and listening exercises is genuinely helpful. I now find that I can do many of the translation exercises solely by listening to the Italian, rather than reading it, which is a good confidence boost.

I was much less successful in my first attempt to learn Italian using an audio-only course. I could parrot particular phrases, but without being able to see the sentence written down, I had no hope of understanding the grammar or even picking apart individual words in the given examples.

Italian is an easy language

I studied French and Latin at school to GCSE level - admittedly a long time ago. Italian's vocabulary is very similar to French and is, of course, ultimately derived from Latin. And, barring a few "false friends" and irregular constructions, Italian often frequently overlaps English. So it's actually a pretty straightforward language for me to learn. I am under no illusions that this is typical; a language such as Japanese would be at least an order of magnitude harder.

I'm still not fluent (and probably never will be)

Duolingo includes a score of fluency, but even Duolingo's biggest fans ridicule and ignore it. I am at a score of 42% but, with a vocabulary of less than 2,000 words, I do not believe this has any real-world meaning. In point of fact, I very seriously struggled to converse in Italian during a recent holiday.

This is likely to be a significant problem with learning any language without interaction with other speakers. Identifying and translating short sentences, even ones with idiomatic meanings, are far easier in isolation than in the context of an actual conversation.

There's an Internet connection everywhere and finding five minutes each day is easy

I have maintained my year-long "streak" despite some pretty significant life events over the past year, and despite being stuck in some unexpected places for both work and leisure. However, the successful streak means that I can say with certainty that I have had daily access to a WiFi connection, and that I have had my tablet to hand for at least five minutes every single day.

I may not be as wedded to an always-on Internet as some in the Millennials generation, but nor have I been without it for more than 24 hours.

I'm unbelievably risk-averse (or the app doesn't understand my motivations)

After a recent update to the app, I am now offered a daily wager: bet Lingots against maintenance of a week-long streak. On the face of it, this is a great deal - I'm already very highly motivated to maintain my streak, so this should just be a guaranteed little bonus. Yet I never take the bet.

We could speculate that I have an exceptionally low tolerance for risk. But actually, the truth is, I don't care at all about winning a few extra Lingots. My motivation lies elsewhere. I think it's great that Duolingo continues to experiment with different ideas, but not all its experiments are successful. I don't think that this wager adds anything to the learning experience. Similarly, there is now a tendency to interrupt lessons with a screen that is intended to be motivational ("Three in a row! Well done!") but actually this just gets in the way; I expect this feature will be quietly dropped in a future update.

It's almost impossible to learn when tired

It's obvious that our mental faculties decline when we are tired, but learning with Duolingo has provided me with solid evidence of just how useless I am at the end of a long day. When tired, I will typically elect to do a practice lesson rather than study new vocabulary. Generally, if I undertake a practice lesson before about 9pm, I can complete it with few or zero mistakes. But if I leave it just a little later - 10.30pm, say - then I might make errors on more than half of the exercises. Some exercises I will get wrong even after being shown the correct answer.

This trend is so stark that it has me wondering about other activities that I might attempt when tired. For example, is it safe to drive if my error rate has gone up to this extent? And beyond what point is it counter-productive to stay late at work?