In which I explore why we want to believe in magic.

Here’s the transcript from this week’s Email from Abroad for The Guardian’s Tech Weekly:

Those who’ve been listening to Tech Weekly for a while know I have a certain level of resistance to apps and services that can be lumped into the thematic bucket, ‘digital serendipity solutions’. These are computerised systems that say they can bump you up against something that might just change the course of your life. But full disclosure here, my experience of them has been largely theoretical, because there have been very few things of this type that do anything other than put you in the path of random information. Until now.

For the last month, I’ve been on the road around the world, and rather than use my trusty Lonely Planet or Frommer’s guide, I’ve been letting the internet guide me. I’ve given myself over to predictive tour guide, Google Now.

Google Now is the search engine’s serendipity engine. It’s their personal digital assistant that, according to the myth, was developed to answer your questions before you even know you have them. The theory goes that the new app will recommend things for you to do and see via your smartphone at the very moment you are ready for them, and these suggestions will be based on where you are, what your search history says you’re interested in, how much time your google calendar says you have before your next appointment, and other data scraped from your digital universal google profile.

Which is how, perhaps I was directed on my first night in Auckland, New Zealand to a local oyster bar (I was raised in the seafood hub New Orleans, according to Wikipedia) within walking distance (a mode of transport I invariably select when using Google Maps) with a fine selection of local craft beers (I make it a point to find the local brewery tour in most places I visit, but I didn’t know Google knew this), frequented by young professionals in their 30s (I am). Earlier that day, I’d been to a vegan cafe, at Google Now’s suggestion, probably because it knew I’d spent time in LA, and there are more vegans per square inch than grains of sand under their recycled flip flops.

Unsurprisingly, the oyster bar was a gastronomic hit. Score one for Google Now. But there was another element that made this experience even more delightful.

Over the appetiser, I happened to start a conversation with a couple who were also just stopping through town, dropping off her mum at the airport and catching a comedy gig in the big city before heading back to their farm on the coast two and a half hours away. Through the briefest of small talk openers – what do you do etcetera – we discovered we had a friend in common. A friend from Canada who now lives in London.

That was weird. Very good, Google Now.

A week or so later, I was in another restaurant recommended by Google Now in another town , and was delighted to discover that they were serving the same bottle of wine I’d almost bought a few weeks earlier when I was in another hemisphere entirely, but hadn’t. Coincidence? I was certainly looking for one.

It was the beginning of a slippery slope. In the old days, my experience of a place would have been guided by conversations or the options laid out in an index. Now, I was looking at my phone to tell me what was nearby. And why? Because it is my externalised brain, not a person who can only guess what I might like, or a book that only dishes out generic advice to independent backpackers. My phone is a simulacra of my interests, my desires and my timetable. And to ensure a hit, it needed more.

And so I migrated my calendar to Google so it would know what my schedule was and where I’d be, and with whom. I uploaded information to my advertising profile so it would accurately place me into a demographic silo. I sent emails when I didn’t need to, just to get it to pay attention to the keywords I dropped in there. In other words, I fell – hook, line and sinker – into its trap. Google Now is a gateway drug to get me to integrate all the loose ends that didn’t already belong to the Google Panoply.

Still, after a while, its recommendations started looking the same. It mostly continued to try to get me to eat, and eat lots, revealing its relationship with Zagat, the restaurant review bible it bought in September 2011. I found myself trying to train the thing: I searched for every variation of the words “run”, “running”, “jogging” to get it to learn that I had brought my trainers to the other side of the world for a reason, but it never recommended a local route. It only recommend restaurants, and pricey ones at that.

I started to lose faith in the machine. I started to question why I should be more willing to believe a device rather than a non-digital curator of local culture. I went to restaurants that weren’t recommended, and had a delightful time anyway. I started to use my eyes to navigate the city, rather than keep them glued to Google Maps to get me to the place that I believed I had to go to have the best experience possible in the few days I was going to be in a place I’d never been before. I’d found the end of the yellow brick road and I could see that this delightful device was just a man behind a curtain.

Google Now is still a work in progress, and the wall I’m now pressed up against is a feature request. As it currently stands, it’s likely that everyone’s Google Now experience is overwhelmed by photos of local landmarks and restaurant suggestions because that’s all they’ve got in their system.

Life is more than just exchange rates, weather forecasting, public transport and entertainment. The good news is that almost every week they have a new card in their deck. But like the characters in the role playing game Deus Ex, Google Now is at a branch in its character path. In that game, the events adapted to the special skills that the player developed: a sharp-shooter played Deus Ex like a first person shooter, while a logical thinker played it like a puzzler. But these simplistic – yet divergent – paths were laid out by the designers based on the virtual bricks and mortar they programmed into the relatively closed system, and their boundaries were very strict.

So how many parameters can Google Now actually take before it’s algorithm melts? That’s a question faced by all recommendations services: what’s the bare minimum of inputs that need to be integrated before its outputs become useful to the person on a work trip and someone on vacation and someone in the office and someone at home? Google Now’s challenge is to work for all of these people, and for everyone else too.

My little experiment raised another question: what is the threshold of things a system needs to know about you so the smoke and mirrors don’t dissipate with the gentlest of breezes? What is the bare minimum to give just enough to fool most of us? And who’s motivation is that, anyway? Unlike the Wizard in his mythical land of Oz, Google isn’t in the game to fool us, so their task is easier than that poor soul’s. They don’t want to be magical. We want them to. Which says more about us and our relationship with technology than Google’s intentions with their new toy.

The terrible truth is that Google could never have predicted the strange coincidences that happened early in my journey. The wine in the cafe in Melbourne, the vegan restaurant in Auckland, the strange connection in the cafe. Yes, they were experiences that were perhaps remarkable to the uncynical eye, but really, each of these things could be explained. Even the mutual friend I shared with the random Kiwis in the cafe. I fell into a pattern recognition trap, I was looking for faces in the clouds. I was delegating human psychology to the magic machine. Google Now was laying the seeds for my personal Barnum Effect, that delightful psychological trick played by fortune tellers from here to Taiwan: it’s recommendations were vague enough to feel personally applicable, especially if I thought it held the answer.

One of these days, with enough cards in the pile and the brain power behind it, my Now recommendations will be different from everyone else’s, just like my search results currently are. The way it delivers both of these things is science, not magic. It’s an algorithm, not a view into my psyche or the fabric of the universe.

Now, though, as in “in the present”, I’m going to put my phone down and go for a run without a map to guide me. Crazy as it feels to leave my digital equivalent of Luke Rheinhardt‘s dice in the hotel room, I’m actually delighted: I have no idea what I might discover.

Listen to the audio on Tech Weekly