I have just read a fascinating article describing how political campaigns, most recently the Trump campaign, use the internet, social media, AI, and related technologies to target and influence voters. Assuming its account is accurate, what it is describing is what political campaigns have long done, just doing it better. It is a normal part of campaigning to try to target ads and campaign speeches to particular groups of voters, telling each group what the campaign believes its members will find most convincing. The difference is the granularity.
This time they have it down to the individual voter.
The first step is to get information about voters at the individual level. Facebook provides one way of doing so, by looking at what people and pages each individual likes, along with the information individuals report about themselves. From that, deduce as much as you can about personality, beliefs, all of the things that will make one argument more persuasive than another. Target each individual with the campaign ads you think will persuade him. Watch what ads each person clicks on and use that feedback to steadily refine the model, improving both your information about what sort of person he is and about what sorts of ads work with that sort of person.
That approach to persuading voters is no more dishonest than conventional campaigning, just better done. Indeed, one might argue that part of what it is doing is reversing the effect of earlier technological change. In the world of 19th century whistle stop campaigns, where most of what you knew about a candidate was what you or your friends had actually heard him say, candidates could and did tell different stories to different groups of voters. The rise of national media made that harder to get away with. The speech telling farmers that, if elected, you would do all you could to keep crop prices high might be less persuasive to urban voters who read about it in the newspaper, heard it on the radio, saw it on television.
But now, through the power of newer technologies, you can not only tell farmers you are particularly concerned with their problems without anyone else overhearing the conversation, you can target your ads to sugar beet farmers in North Dakota. And you can vary the particular version of the ad according to what sort of argument each individual farmer finds most compelling, as reflected in which ads he reads and shares as well as the information deduced from his other online activities.
So far, it is merely conventional targeting done better. The anonymity of online conversation adds another tool for persuasion—fictional identities. The technique is familiar from online arguments where three of the four people supporting a position may be sockpuppets of the fourth. Done on a professional scale, the hundred people arguing for a position, insulting opponents, praising supporters, may be three people in one office.
It is an interesting article and worth reading for a more detailed account of the technology. The author obviously has his own axe to grind, since he believes that at the moment the leading practitioners of the new dark art are working for the wrong side; he makes what he is describing sound more sinister than it actually is, most of it being simply new and better versions of ancient tricks.
As he makes clear, it is a technology that could be used for any political cause. It could also be used for non-political causes, and is. Presidential elections are every four years, but people are always buying and selling. Technologies that tell a political campaign what arguments will persuade a voter to vote for their candidate can as well be used to persuade a buyer to buy what the seller is selling.
Seen from this standpoint I find the technology considerably less sinister. It is true that it can be used to produce bogus praise for products from fictional users, but I already knew that and discount such praise accordingly, more heavily the more it feels like something that could have been churned out by a computer program. But it can also be used to try to sell me the sorts of things that I want to buy. I would much rather receive emailed ads for opal rough on ebay—one of my hobbies is lapidary work—than be interrupted at dinner by a phone call from someone trying to persuade me to refinance my nonexistent mortgage.
"It is true that it can be used to produce bogus praise for products from fictional users, but I already knew that and discount such praise accordingly,"
Yes, David, you, and I, know this. But many people with less education and less knowledge of what can be done with computers may not. I am very leery of how that kind of tool could be used by a very few people to sway many, and I don't know what the best way to address that problem is.
Part of the answer, as with ordinary political campaigning, is competition. If the Republicans are trying to persuade us for one thing, the Democrats will be trying to persuade us against it.
More generally, I think ordinary people already discount advertising--not entirely, but to a considerable degree.
Talk about click bait. I was expecting to read about a new Japanese sex robot.
In this connection, you might be interested in this account of the Vote Leave (Brexit) campaign manager: http://blogs.spectator.co.uk/2017/01/dominic-cummings-brexit-referendum-won/
It's quite long, but I found it fascinating.
"It is true that it can be used to produce bogus praise for products from fictional users, but I already knew that and discount such praise accordingly"
Maybe you do, but the search engine algorithms that decide what you see and in which order don't do this discounting, hence why such techniques are more sinister than at first glanc.
Ok, I tried to read the article, but it's on an invitation-only website with a "waiting list" to sign up. Am I doing something wrong? I don't want to get on yet another mailing list, and doubt that there will be many articles there I would want to read anyway. Is there a way you can provide a link directly to the article, or post a copy of it? I'd like to read it, but not enough to create an account there.
Another fine reason to stay off Faceplant and Twitter.
Have you read the book Lexikon, by Max Barry? It's about exactly this sort of individualized marketing, but done mostly by individual humans. (In the last chapter there's some reference to a new wave of tech geeks getting into the game, analogous to the distinction in stock markets between fundamentalists and technicians.)
Anonymous's point about search engine algorithms is important: it's hard to discount an opinion when you don't even see the contrary opinion. I wish there were a way for an individual user to program the search engine algorithms, e.g. to say "yes, I'm politically liberal, but I honestly want to see politically conservative content." I suppose one could try to jigger the system by intentionally clicking on things that one is likely to disagree with, whether one has time to read them right now or not, but it'll take constant vigilance.
here's a recent article suggesting that Cambridge Analytics' claims about winning the election for Trump are rather exaggerated, https://www.buzzfeed.com/kendalltaggart/the-truth-about-the-trump-data-team-that-people-are-freaking?utm_term=.vw5Ymla8O#.so4ayJoq2
Interesting, but hard to tell how much substance there is to the criticism.
"[W]ithout anyone else overhearing the conversation" looks like an exaggeration. Your political opponent can rather easily get a hold of a lot of your targeted messages--they can't be *perfectly* targeted, nor can you stop them from being passed on to untargeted recipients; these messages can be assembled (for publication) into a composite which plainly shows your hypocrisy.
Of course computer and Internet are future of the world. Trump did an amazing work: he used a media which is impossible to be manipulated so he won election :) I think TV and other media than Internet will be dead in maximum 20 years.
You explained the reason why i don't use twitter .Thanks man.
Post a Comment