Can Computers Make Better Social Media Judgements Than Humans?

Posted on 19. Aug, 2011 by in Digital Marketing, Thoughts



Computer and manThere has been a lot of discussion about services such as Klout.com and PeerIndex.net recently. I’ve read strong opinions that online influence, authority etc. can only be accurately determined by an experienced social media expert and not algorithmically. There is evidence to suggest that this assumption is not a fact and indeed the opposite is true.

Are computers actually better at judging people online than humans, even social media experts? If not, can or will they be in future?

The human expert views a very limited section of the web but may be able to glean some human insight from the limited number of interactions, data points and content they are able to collect and investigate. However, judgements are likely to be inconsistent and partially based on intuition.

On the other hand, computers are capable of aggregating and analysing unimaginably large sets of data from billions of social profiles; forming an infinitely complex social graph over numerous platforms. They can investigate how profiles are related to and interact with each other and how they are related to and interact with content. They can examine all content created by a profile and see relationships and interactions over time. Algorithms can be built which quickly and easily identify profiles with specific characteristics that might make it likely they are, for example, an authority in a certain field, a profile who interacts a lot or a spam profile.

How far can we take it?

Perhaps it is also possible to distinguish personality traits or predict future behaviour algorithmically.

Admittedly, public systems such as Klout are nowhere close to being able to provide us with that level of insight yet but three years ago university researchers discovered that narcissism was a trait easily identified by examining Facebook profiles. I wrote this at the time:

Looking at the wider implications of this research, if it’s possible to spot this personality trait so easily, presumably it could be possible to spot others too. Could it be as easy to tell is someone is an introvert or extrovert, neurotic or pervert just from their Facebook profile? If it is, you can bet someone somewhere (probably the social media sites themselves) are building algorithms that work it out automatically in order to direct better targeted advertising at users.

It turns out those words were correct. Earlier this year researchers at Maryland University published a paper on accurately predicting users’ personalities through the publicly available information on their Facebook profiles.

It’s probably best if you go back and read that last sentence again. It is possible to accurately predict people’s personalities from their Facebook profiles. Wow! The potential for personalisation and marketing is huge, albeit the ramifications are also quite worrying.

Learning from the mistakes of others

It is understandable that someone would think judgements such as those made in social media must be made by humans rather than crunching numbers. By their nature, social networks are about human interaction. However, this assumption has parallels to those wrongly made in other fields.

In his book Supercrunchers Ian Ayres mentions a number of examples of relatively poor judgement by human experts when compared to computers crunching mountains of data.

Years ago, getting a loan or mortgage from a bank involved putting on your best suit, combing your hair and going to meet the manager. Your chances of getting the loan were affected by your relationship with the bank manager. It wasn’t purely based on your ability to or likelihood of meeting the payments.

It turns out that computers with access to all your background data and importantly, with no emotional connection to you, are far better at judging whether you will default on the loan. The result is that the power to make these decisions has been removed almost completely from branch staff. They simply weren’t as effective.

Little Britain, Computer says no

Little Britain - Computer says "no"

Ayres gives numerous other examples where data outperforms human knowledge in areas where we naturally assume instinct, intuition, intellect, experience and other human qualities are required. Areas such as predicting the best vintages for wine and evaluating baseball players. Both of these jobs have been completely transformed since they figured out that data trumped intuition.

So where does this leave social media experts?

Social media experts will always be required, it’s just that the talents they should be sharpening involve putting together solid, imaginative marketing strategies, running campaigns well, making good PR decisions, forging relationships and a million other creative things where computers will (probably) take a long time to replace people.

Tools like Klout should not be ignored, even by experts. They might not be particularly advanced yet but they do a decent job of quickly finding out a lot about a profile.  Publicly available tools have not yet reached a point where they are complex enough to provide an analysis that doesn’t require some further digging or backup. However, the information they provide can be highly insightful since given the time constraints, humans could not possibly collect and interpret the same amount of information consistently or accurately.

Eventually, computers will almost certainly become more reliable than social media experts at analysing and judging people online. The marketing implications for social networks of being able to predict their users’ personalities are too huge for them to ignore.

Richard Falconer

Tags: , , , ,