Hacker News Reader

Comment by Chickenality | Hacker News Reader

For those who find this horrifying, the [typically unstated] assumption is that it works. I'm pretty skeptical of this. Think about it: what do you imagine the effect size is of seeing 10 articles like "4 Secrets to Losing Weight and Feeling GREAT While You Do It!" in your Facebook feed? I'd bet it's essentially 0. If seeing these kinds of articles is all it took to improve our body composition, then we'd all be absolutely shredded.

I'm not trying to argue that advertising doesn't work. Clearly it does. But there's a huge difference between being exposed to tens of thousands of Coca-Cola ads over a lifetime, and being exposed to 10 spam articles over the course of a month.

I'm also not trying to argue that this isn't creepy. Clearly it is. But the real 'target' of this scheme is the person buying the ads. I doubt this has any real effect on the person who sees them.

View on HN

For some scientific research into this subject, see:

https://en.wikipedia.org/wiki/Priming_(psychology) https://en.wikipedia.org/wiki/Repetition_priming

(do note that this research suffers a bit from p-hacking).

There is a good chance that you were already targeted before, since online manipulation is used by the big militaries. Think back about 2 years, reading about SJW, politics, neo-nazis, antifa, BLM, manspreading, immigrants, the deep state, etc. Good chance at least some of your perception about these subjects was molded by just a few individuals. For instance, remember that Russian girl throwing bleach on "manspreaders" in the metro? You may have had a strong reaction to that, and it would be the desired effect of Russian troll-factory.

Specifically, on the effect of manipulating the Facebook feed to control behavior, Facebook did some controversial research themselves, where they used sentiment analysis to make a feed more or less positive. People who were fed negative feeds, started using negative words in their own status update: https://www.forbes.com/sites/gregorymcneal/2014/06/28/facebo...

> The marketing study suggested companies should "[c]oncentrate media during prime vulnerability moments, aligning with content involving tips and tricks, instant beauty rescues, dressing for the success, getting organized for the week and empowering stories... Concentrate media during her most beautiful moments, aligning with content involving weekend guides, weekend style, beauty tips for social activities and positive stories." The Facebook study, combined with last year's marketing study suggests that marketers may not need to wait until Mondays or Thursdays to have an emotional impact, instead social media companies may be able to manipulate timelines and news feeds to create emotionally fueled marketing opportunities.

Long term priming, as opposed to short term with gaps of a second or so, was one of the prime victims of the replication crisis and seems to actually not exist.

Priming is one of the psychology topics worst afflicted by p-hacking. Almost every prominent priming study has gone down in flames during the replication crisis. Consistent, repeated propaganda has real effects, priming doesn't.

Like said by the other poster, long-term priming is not proven nor disproven: Science needs better and more experiments. So we don't yet scientifically know enough about long-term priming to make a judgment on its effects. Short-term priming is well-established and has real measurable effects though. The prominent priming studies you refer to are the "exotic" studies -- these looked at less defined aspects of priming, and were found to be lacking.

> Amidst the recent furor over failures to replicate some empirical results on behavior priming, it is important to emphasize that some basic behavior-priming effects are real, robust, and easily replicable even if others are much more problematic.

For instance, your reply contains too many words starting with "p" and "pr" for it to be a mere coincidence :). (syntactical priming is something that authors or editors have to guard against, as it can make for poor quality writing).

> Like said by the other poster, long-term priming is not proven nor disproven:

If there is no evidence for it, there is little reason to be concerned about it at this point given how much research has gone into it.

Absence of evidence is no evidence of absence. You could put the Bayesian prior to be extremely low, but a zero chance would not make you a Bayesian anymore, it would make you a believer in that something is simply not possible (and no amount of scientific evidence would update your priors. It really is scientifically a mistake to claim: There exist no black swans. To proof that, one would have to observe all of existence. Now... should you worry about black swans, when all you see is white swans? Depends on you and the amount of risk managing. But that poster claimed all of priming is non-scientific, when we have clear replicated proof.

Suppose I run a fake investment company that pretends to double your money, but actually just steals it. Suppose you find out all my claims are lies, and demand your money back. How would you react if I said "The absence of evidence is not the evidence of absence. The prior probability of my investment strategy working is nonzero. Now give me more money."

Just pointing out that priors can't be zero isn't a principled stance, it's Pascal's mugging. You can use it to justify literally anything.

>Absence of evidence is no evidence of absence.

No shit. But at this point it’s pretty safe to not waste money building anti-aircraft guns to shoot down flying reindeer.

The prior for priming effects as strong as those claimed by psychology is, obviously, extremely low. After all, the only reason everybody knows about priming is that it got lots of press coverage, because it was surprising and unexpected. Now that we know the evidence for it is very weak, we can just go back to our original prior for it, which was very low.

In general this is true for almost all popular psychology results. They're all popular because they're surprising, which is what makes them interesting. And now we are learning that such surprises were only produced in the first place because the evidence for them was p-hacked. It turns out that there are no tiny hacks that radically change human behavior, beyond placebo effect.

I'm sure that psychology produces some real results, but whatever they are, they aren't what get reported in the New York Times, or TED talks, or bestselling pop science books.

It might have a low chance of success on one particular person ... but target ten or a hundred thousand, and there will be some statistically significant shift. Advertising isn't a sniper rifle. It's a cluster bomb.

You have to remember the majority of people are both not tech savvy and don't pay attention to things like advertising trends.

My mother for example is heavily affected by the ads she sees on Facebook and other places. All it takes is one or two articles and I've had to talk to her a few times about being careful about the sites she visits and what she reads online.

So this service? This would be remarkably effective on people like her. This service isn't for people like us who are a bit more aware of how advertising works. It's for the majority of people.

Yeah, it comes off as more as a way of taking advantage of desperate people willing to try anything.

Isn't this what enabled the Brexit?

I mean, you can show people ads about "Are you an immigrant? Claim your government-sponsored house. We will help you to qualify even if you don't meet the criteria" and now you conceived the idea that the government is giving away free houses to immigrants and there's even a way to cheat the system and get a house even if you do not qualify.

As far as I remember they essentially divided people into categories through data gathered through facebook and targeted them with ads like this.

I guess if you can manage to put the cookie to identify your target you can run all kind of manipulative campaigns because you can change the perception of the reality.

You don't have to be honest and straightforward and speak directly to the target. Simply re-shape his/her reality.

Want a rise? Stalk your boss with ads about developer jobs with salaries way higher than yours. Your boss might start getting the idea of pay rise due to the false reality you prseented.

quote: 'I mean, you can show people ads about "Are you an immigrant? Claim your government-sponsored house. We will help you to qualify even if you don't meet the criteria" and now you conceived the idea that the government is giving away free houses to immigrants'

I think I actually saw an ad similar to that - my takeaway was "damn, these slimy scammers are targeting desperate immigrants", so I guess that means they didn't hit their target?

The inability to quantify targeted advertising's worth at specific levels is probably seen as a benefit by those employed on both sides of the transaction (i.e. marketing side and serving side).

My apologies I'm using an app that is not intuitive to comment and read.

Reply on news.ycombinator.com