杏吧原创

Editorial: 鈥�Unmasking hidden persuaders is a hard sell鈥�

I first realised I had a problem inside the department store. I was meant to be choosing a gift for my niece. My bored 2-year-old had begun pulling the items off the shelves, in a desperate bid for entertainment. That was because I had been standing still, just staring at a display of colourful children鈥檚 toys, for well over 10 minutes.

The problem was that I had left my smartphone at home. Usually when I can鈥檛 make a decision, I whip out my phone and ask Google or one of my apps for advice or reviews. Without it, I was paralysed by choice.

I now continually turn to apps and websites for recommendations about what to buy, what to eat and where to go. Life鈥檚 big decisions remain my own, but many of my everyday choices are essentially delegated. As well as regularly turning to search engines for advice, I use apps like MapMyRun to set a jogging route, for instance, and iTunes to select the music to listen to while I run. I let Google News tell me what stories I should read, and have Netflix pick which movies to watch.

I鈥檓 not alone. For example, more than one-third of on their smartphones for guidance just before they make a buying decision. Smartphones and tablets have become more than simple communication devices: they are very often the source of the information we use to decide how to navigate the world.

Although there is nothing wrong with being constantly plugged in and informed, this habit inevitably means that our everyday decisions are being subtly influenced without us even realising. You might not have elected to cede control of your choices, but if you spend any time online, it is almost certain that you are unwittingly doing so already. And in the next few years, we could hand over even more autonomy. Our portable devices are poised to become intelligent 鈥減ersonal assistants鈥� 鈥� perhaps even capable of acting independently on our behalf in work, entertainment and relationships.

So what are the implications of delegating everyday choices? If we are not doing the deciding, who or what is? And should we be concerned, or might removing some choice from our lives actually make us happier?

We all have free will, and it would be ridiculous to argue that turning to an iPhone for advice could take that away. Still, we can only make informed choices on the basis of the information available. A lot of the time, that information comes from a product review usually written by a person. But increasingly, much of the information and recommendations we get from the internet have already been filtered and personalised by 鈥渞ecommender鈥� algorithms. These bits of computer code use our previous preferences, and those of people with similar tastes, to present a view of the world that is different to that of our neighbour鈥檚.

For example, websites like Amazon and Netflix continually recommend items targeted at you, and many smartphone apps now do the same. One app, called , uses your past choices to suggest nearby restaurants you鈥檇 like, and is dubbed 鈥測our personal robot鈥�. These algorithms are already highly influential. Around 60 per cent of all movies rented from Netflix are based on the recommendations of the site鈥檚 algorithm, says the company.

Since 2009, Google has worked in a similar way, personalising the results provided by its web searches for each individual, largely based on their past activity using the search engine. Social network feeds are also personalised: the status updates and photos that appear on your Facebook news feed have been edited by an algorithm based on factors such as how it rates the strength of your friendships (see 鈥淕ot friendly world syndrome?鈥�).

By carrying our tastes, previous choices and background around with us as we visit each corner of the internet, we are enabling companies to make decisions on our behalf about what we see when we are there 鈥� which in turn will influence our subsequent choices.

For an illustration of how, consider this example from Eli Pariser, author of the 2011 book The Filter Bubble. Imagine two people entering the search term 鈥淏P鈥� immediately after the company鈥檚 Gulf of Mexico oil spill. One, a liberal, is presented with news stories about the environmental effect of the spill, while the other is given stock tips about the company. If they are both investors trying to decide whether to buy BP shares, they could come to quite different conclusions.

Even the adverts we see online are personalised, which can affect the range of choices we think are available. As Pariser points out in The Filter Bubble: 鈥淪tudents who go to Ivy League colleges see targeted advertisements for jobs that students at state schools are never even aware of.鈥�

Most of the time all this personalisation is helpful 鈥� especially for battling the modern malady of information overload. Take Netflix. 鈥淚ts users have some version of the entire library of film in front of them,鈥� says who has raised concerns about the influence of algorithms. 鈥淗ow are they supposed to make a decision about that?鈥�

But at the same time, algorithms are far from perfect at forecasting complex human desires, Slavin says. They break down our tastes and behaviour into categories they can understand 鈥� so that because I like Jane Austen novels, I am also likely to watch Downton Abbey, for instance. In reality, we are much less predictable.

Some people might, like me, find it unsettling that all this personalisation is happening without fanfare. Few of us are even aware that it is going on, says Dean Eckles at Stanford University in California, who studies the social influence of interactive technologies. 鈥淚t is basically invisible to people,鈥� he says.

So how concerned should we be? To tackle that question, we need to look at where this technology trend is headed.

After all, the current wave of personalisation is just the beginning, says Justin Donaldson, president of machine-learning company , based in Corvallis, Oregon. He and his colleagues argue that we are on the verge of a new wave of systems that can act directly on our behalf, instead of just suggesting options .

The increasing use of recommender systems on smartphones is giving them access to a rich source of data about our everyday lives. My smartphone, for example, knows not only what search terms I enter into Google and who my friends are on Facebook, but what my GPS coordinates are at any given moment. It also has access to my calendar, emails and daily notifications.

Donaldson predicts that, using this information, smartphones will soon be able to act like personal assistants, if not old-fashioned butlers. 鈥淏utlers knew of all of the quirks, tastes and vices of their employers,鈥� he says. 鈥淭heir task was to promote, suggest or arrange the most appropriate attire, meal or social function.鈥�

鈥淪martphones could become like butlers, knowing all our quirks, tastes and vices, and acting upon them鈥�

For a hint of what is to come, says Donaldson, look at Siri on the latest iteration of the iPhone. Billed as a 鈥渧irtual assistant鈥�, Siri is equipped with the artificial intelligence to understand voice commands such as 鈥渇ind me an Italian restaurant鈥�, and can use the web or apps to search for nearby eateries. According to Apple, Siri also learns about you over time, using information from your calendar, contacts list, music library and reminders to better understand what you mean. It is not impossible that future versions of programs like Siri could use this information to arrange a night out with friends, plan a business meeting or book a hotel for you.

Soon, your smartphone could even make recommendations and choices based on how you are feeling. Hyun-Jun Kim and Young Sang Choi of the in Yongin, South Korea, are developing a system that uses information about a user鈥檚 emotional state to shape its recommendations for music or products to fit their mood. Called EmoSens, it uses the phone鈥檚 sensors to measure signals such as shaking, walking pace and speed of tapping on the screen. It can then conclude that a person is more agitated than usual, for example, and so tailor its recommendations accordingly. For example, if you are looking to book a restaurant, it might propose comfort food if you are in a bad mood. The pair at a conference on recommender systems in Chicago last October.

Do we want algorithms studying our behaviour and acting on our behalf in this way? How can we be sure that a virtual assistant won鈥檛 store our information and sell it to marketing companies, for example, or subtly favour particular brands?

There will always be companies taking advantage of any technology when personal details about our lives are involved. Some people may not like that, but the benefits should outweigh the risks, argues Donaldson. And just because recommender systems today tend to work on behalf of companies that want to sell something, that doesn鈥檛 mean the next generation would do the same, he says.

In principle, tomorrow鈥檚 virtual assistants could work solely in our best interests, not a company鈥檚. So, for example, it might intervene to stop you buying a novel from Amazon, by suggesting you borrow it from a Facebook friend who it knows bought it last month. 鈥淭hese things can genuinely improve our quality of life,鈥� says Donaldson. 鈥淥ur generation just needs to understand where the boundaries are, and how we can protect ourselves.鈥�

Culture shaper

Still, there are wider implications to consider if we collectively delegate more and more of our decision-making to algorithms. The recommender systems used to personalise our entertainment choices could even start to shape culture itself.

It is not impossible that plays and films would make more money if their authors wrote them to satisfy an algorithm鈥檚 idea of what people with particular tastes like to watch. If that sounds unlikely, consider that many online journalists already write news and headlines in a style that ensures Google鈥檚 algorithms place their stories higher in the search results. They are not just writing for their human readers; they are writing for algorithms too.

Another issue also gives pause: if we all have the world personalised for us, what experiences do we end up missing out on? We are often rewarded for spontaneous choices. Think of all the movies you have enjoyed because you took a risk, or meals you liked at restaurants you simply stumbled upon. It would be a shame if the world lost some of its serendipity.

On the other hand, we may all end up happier if we had fewer decisions to make each day. Psychologists know that our ability to make good decisions deteriorates after an extended period of making choices 鈥� even little ones like what products to buy in a supermarket 鈥� in what is called 鈥渄ecision fatigue鈥�. And various studies have shown that people with restricted choice 鈥� or none at all 鈥� often feel happier with a given outcome than those with more freedom ().

In one experiment by Simona Botti of the London Business School and Sheena Iyengar of Columbia University in New York, participants were presented with a meal chosen by a friend or given the freedom to choose their own dish. Often the meal would turn out to be unappetising, but the people who had chosen it themselves enjoyed it much less than those who had it recommended by the friend ().

As someone who hates making decisions, this makes sense to me. Put an intelligent personal assistant in a smartphone, and I would consider buying it. But I would certainly feel better about algorithms acting on my behalf if I had greater control and awareness of when they were guiding my choices. Anthony Jameson, head of the Choosability Engineering research group at the in Saarbr眉cken, argues that software developers should make their recommender systems more transparent. For example, existing search engines and recommender systems could easily display a 鈥渟lider鈥�, from which users could choose how varied they want their recommendations to be, he says.

In the meantime, though, I have resolved to think much more carefully before turning to an algorithm to suggest a product or activity. And that is perhaps one of the better-informed choices I have made this year 鈥� even if it does mean I spend more time staring blankly at shelves of toys in department stores.