back
92 comments
Below is the list of feelings, which, according to the article, the NYT is now predicting with ML models trained on a proprietary dataset collected internally from a sample of articles shown to a sample of readers. I've marked positive feelings with `(+)`, negative feelings with `(-)`, and feelings which could be viewed one way or another, depending on context, with `( )`.

Remarkably, a majority of these feelings are positive, so this advertising initiative, which has been successful at increasing per-ad revenues by as much as 80%, is creating a financial incentive for the NYT to produce content that generates positive feelings in readers!

On its face, this is shocking to me. I would have never, ever expected positive content to be more profitable than negative content! In hindsight, however, after thinking about it a bit, it kind of makes sense. I'm now wondering, could advertising perhaps become (gasp) a force for good, at least in certain settings?

  NYT PREDICTED FEELINGS
  Optimistic             (+)
  Inspired               (+)
  Self-Confident         (+)
  Amused                 (+)
  Adventurous            (+)
  In the mood to spend   ( )
  Love                   (+)
  Sadness                (-)
  Boredom                ( )
  Interest               (+)
  Fear                   (-)
  Hate                   (-)
  Hope                   (+)
  Happiness              (+)
  Nostalgic              ( )
  Indulgent              ( )
  Competitive            (+)
  Informed               (+)
The predicted emotions are actually contained in the HTML source of each article. E.g. for an article I randomly grabbed off the front page:

  curl 'https://www.nytimes.com/2019/04/14/sports/tiger-woods-wins-masters.html' | perl -n -e'/"emotions":"([^"]*)"/ && print $1'
inspired,boredom,happiness,optimistic,not_interest,not_hate,not_amused,not_inthemoodtospend,not_informed,not_love,not_sadness,not_fear,not_hope
> not_hope

This one is a little surprising. Does the NYT think this article dashes any hopes readers might have of winning the Masters?

Is this potentially opening them up to data leakage? Could someone not crawl those pages, analyze the content and use the existing labels as the training set?
It has always been known that certain content would work better to sell ads than other. Travel, cooking, or home decoration, for example, are definitely better than political news.

And yet, the NYT has withstood the pressure to become the ground level equivalent of an in-flight magazine.

Can we maybe credit them for that, instead of just assuming that everybody agrees that they will succumb to economic pressure without feeling the need to even make the case for it?

>* On its face, this is shocking to me. I would have never, ever expected positive content to be more profitable than negative content! In hindsight, however, after thinking about it a bit, it kind of makes sense. I'm now wondering, could advertising perhaps become (gasp) a force for good, at least in certain settings?*

News are good when they are neither positive nor negative by design.

They are good when they are relevant, useful, and informative (including them not being cherry-picked examples of stories with no impact to their readership, e.g. about some sensational murder case 10.000 miles away, and so on).

Feeling good is not the point of news. The point is being informed. I cant help but think of the saying, “ignorance is bliss.”
People tend to upvote a lot more than downvote. I believe it's because one's ego doesn't want to associate themselves with being negative, instead we always try to attach ourselves to more positive traits.

These metrics should only be used for categorizing purpose only. We need to be very careful in using it as a driver for more of the same articles, because in the long run this will only create a feedback loop that will keep feeding us the same content and narrowing our views, making us blind to more critical issues at hand.

More revenue per page view - but its likely that negative or controversial content will get shared more and drive more overall traffic/impressions. So not sure how much of an effect this would have on the overall sentiment of content produced.

https://blogs.lse.ac.uk/politicsandpolicy/why-is-there-no-go...

Manipulating people's emotions to elicit a short term positive response is not necessarily "a force for good."

Journalism has historically prided itself on telling the truth. The truth is not always a feel-good experience.

All good news, all the time is potentially a dystopian scenario where the truth is forbidden because it isn't profitable enough.

i m willing to bet that 80% more revenue is due to the novelty factor. i believe the ad business is extremely more arbitrary than it is thought to be, and advertisers might be buying these slots because they are new and exciting, not because they are more effective.
This is just their feature space. It's just what they use to distinguish between articles. It says nothing about their incentives.

Also, competitive as positive? I'd see that much more as a neutral (or maybe negative).

Negativity is for clicks, positivity is for renewing subscriptions.
I would add that the NYT is a very niche case. Their target audience has been educated coastal liberals for some time. It may be that this is a result of their audience and not predictive of an underlying phenomenon.
I can see the headlines now:

  "Thousands killed in [less-developed country] civil war; more food left for survivors!"

  "Average global temperatures up 4' C in past decade; boom in tourism in Hudson Bay and Siberia!"

  "BigCo lays off 70,000 workers; social services employees self-worth increases!"
It would seem that such ad placement, based only on content and not the reader, is exactly what privacy advocates have been clamoring for.

I know the mention of emotions creeps people out. But it’s probably the only salient connection a general interest publisher such as the Times can offer advertisers. Very few of their articles mention any product, product category, or really any factual nexus to anything somebody might want to sell.

This sets up a horrifying set of perverse incentives for journalists. This is not to say that the set of incentives facing journalism today isn't already pretty bad, but this forces entities like the NYT further onto a path of convergent evolution with the National Enquirer.
Sounds like a great conflict of interest. Emotional manipulation will become a metric to optimize. Depending on what the content manipulates you into feeling, you may sell more ads. This leads to more manipulative content since it's what sells more ads.

As an alternative to current targeted advertising, this seems much worse. At least I can block ads and trackers. I can't block content that's been engineered to manipulate.

I agree.

From the article - """Journalists and other news purists may also wonder whether the Times newsroom is now being tasked with producing lots of emotional stories.

Neither is the case, Allison Murphy, senior vice president of advertising innovation, told me in an interview."""

If this isn't a lie already it will be in a few months.

The simplistic counterargument is that writers do this already but as we know... once we have numbers to target the optimizations will kick in.

This won't end well :P

doesnt bloomberg already reward articles that shake the markets? (at least thats a rumour)
"The offering also includes “neutrality targeting”: that is, isolating upsetting stories that could decrease receptivity to ad messages — placements to be avoided."

In a news environment driven by ad revenues, expect the news to avoid hard hitting stories that make you feel bad. Shame and anger are emotions that cause change -- they don't sell consumer products well. Chipper stories are the order of the day, never mind the complete capture of the information system by corporations, government elites, and the rich generally.

News environments have been driven by ad revenue for over a century. Serious newspapers address this by a Chinese wall between editorial and advertising departments [1].

[1] https://en.wikipedia.org/wiki/Chinese_wall#Journalism

You should maybe coordinate with the other popular conspiracy theory, namely that „the media“ is driving outrage because it sells so well.
Your idea that they'll avoid articles that make you feel bad is incorrect.

> placements to be avoided

Doesn't mean they won't have advertising on them, just that it will be slightly cheaper. Everyone wants to advertise in the NYT but it costs too much. If I get my ad there, but slightly cheaper, I absolutely would.

> Shame and anger are emotions that cause change -- they don't sell consumer products well.

I don’t know, news has been ad driven for a long time and is notoriously negative.

I know I live a comfy life but the disconnect between life and news is staggering; consuming news seems to me to be horrible shit nearly 24/7, almost all of which is completely absent in my day-to-day life. And yet almost everyone I know — comfy people like myself — consumes tons of news.

Could be that advertising does better per-reader on positive stories, but doesn’t bring readers back as well as frightening news.

I wonder how much they’re hacking the “Popular” section in the app to increase ad revenue. I notice that absolutely 0 of their top stories are ever at the top popular article list - usually travel, opinion and health, which presumably has more high traffic ads.

They’re also messing with the scroller - it will stop scrolling on an ad to make sure it loads.

Here's a video from one of the data scientists who worked on the project explaining the science behind it: https://www.youtube.com/watch?v=sVzZmNxmWko
Good! I can't wait till I have a browser plugin that will tell me what my mood will be after reading a news article. You could make it an add-on to RescueTime. One could limit depressing or enraging news articles to 10 minutes a week.
NYTimes is the main reason I've been on the search for some kind of feature for HN (addon? greasemonkey script? ANYTHING) to auto-hide articles based on domain or title matches. Does anyone have a solution?
> NYTimes is the main reason I've been on the search for some kind of feature for HN (addon? greasemonkey script? ANYTHING) to auto-hide articles based on domain or title matches. Does anyone have a solution?

You'll leave an orphaned article info table row with this approach but you can hide the article row itself with uBlock Origin procedural filters[1].

Add the following to 'My Filters':

news.ycombinator.com##table tr tr:has([href*="nytimes.com"])

[1] https://github.com/gorhill/uBlock/wiki/Procedural-cosmetic-f...

Ohh this brings me back to my Stackoverflow days. HN simplicity makes this relatively easy, here you go (greasemonkey): https://gist.github.com/m00g00/99084d0ca401ab411e518bfdee46a...

It's just a domain search, but should be easy to modify for title searches too (title is in a.storylink).

EDIT: Updated to filter on keywords in the title too...guess I'm bored: https://gist.github.com/m00g00/e539ec22bf588edca0e6dfe1a05eb...

It is difficult for me to see this as not being naive. Yes, the editorial staff of any newspaper would have opinions, and even worse, implicit biases. 1: that does not mean that the firewall between journalists and editorial is being broken casually; 2: you correct for bias by using diverse journalistic sources (and removing only the worse offenders like say Breitbart or Infowars), not by completely disregarding information from imperfect sources.
On one hand, I like this because it might skew "news" production toward positive news stories. On the other hand, I dislike it because the news should just be the news and the business of the news should not skew the content of the news. But, that's probably too much to ask these days.
Bet the Financial Times doesn't do this.

(See https://www.cjr.org/special_report/why-the-left-cant-stand-t... -- whatever your politics she makes a fair point)

Is this one of those instances where regulatory laws would help? Something like requiring publishers to place a notice saying, "This article typically invokes hope. The ads shown on this page utilize this tendency".

These premium ads look like the logical progression of content-based advertising. The only slimy thing is them pretending to be randomly placed.

Most respectable brands conform to AdChoices policies. If you click the little blue triangle on an ad it will usually tell you how an ad was targeted to you.
If there was any question that the NYT is not a newspaper, at least in the pre-internet sense of the word, this should put it to rest. The same kind of shenanigans are at play in every news outlet I can think off. How does one make sense of such a vast world when telling the truth is not profitable?
> telling the truth is not profitable?

Where'd you get that idea? Plenty of newspapers tell the truth and still make money. They do make a lot less than previously though.

Why is the advertisement industry so sleezy and disgusting? Aren’t there any sensible ways of putting your product or service out there without insane advertisement strategies that invade privacy, cross ethical boundaries and sometimes, literally illegal?

Advertisement is why we have privacy issues (the other being governments spying on their people) and giant corporations based on targeted advertisement business model. It’s sickening and people working in this profession should be ashamed of themselves. Not to mention all the psychological tricks that go into the “smell” of the store, deluding people into buying and then underdelivering, buy one get one free, sales!!! sales!!! sales!!!, 20% off if you sign up for a store credit card, etc. what happened to ethics and honesty in doing business?

There has to be a more sensible way? Word of mouth, magazines, banners, trade shows, demos, putting product on the shelf, etc aren’t good enough? I am not a sales man but boy I despise that whole industry.

Is this really more slimy than targeted ads? At least this is based on the content, not the person's information who is browsing it.
One of the dead comments in this thread calls for "criminalizing advertising". I am quite positive to this idea myself. Does advertising really do anything positive to the economy and the world? Doesn't the paradox of advertising already prove that it is at best unnecessary?

Recently however I read about the time in the middle ages when guilts had forbidden advertisement. They took this to such a degree, that if a merchant would cough while a potential customer was walking nearby, the other merchants would accuse him/her of trying to catch the attention of the passer-by.

Perhaps the reason we allow advertisement is just that we don't know how to get rid of it. However, then we should certainly try to regulate it as much as possible; make rules limiting it to the tolerable and perhaps undamaging.

Unfortunately, it may be that so many industries now depend on selling adverts, that limiting it in any meaningful degree would be very hard politically.

All industry will be sleazy and disgusting if it pays to be that way. If giving kids cancer meant they could make an extra dollar, they’d do it. (And there are plenty of examples of other industries doing exactly that.)

The only way to put a stop to it is to make it unprofitable.

There was a story a few years ago about a girl scout who found that she could be incredibly successful by setting up shop outside marijuana stores.

From one angle, this girl is giving the customer what they want. It's a win-win.

From another angle, this girl is using psychological tricks to exploit her customers to convince them to buy eat high-fat high-sugar cookies.

If it were easy to come up with a clear boundary between ethical and unethical advertising that most people agreed with, I believe we would do it.

There has to be a more sensible way?

Like what? If you were to put forth a recommendation it actually seemed like it could work, that would be something people could go forward with. Short of that, you're asking companies to not take advantage of legal methods that their competitors surely will take advantage of.

But there are no privacy concerns in this case. The ad placement is completely contextual based on content, not based on viewer. They just trained a model to learn what sentiment an article would elicit and allows advertisers to target based on that. This seems like the best case for advertisements for anyone concerned with privacy and targeting.
Competition. Other companies do shady stuff (that works), so you "have to," too.
What's slimy and dishonest about having a store that smells nice? Do you object to cleanliness and a pleasant temperature too?
This is not particular to NYT but when your news sources are A/B testing and have information on your feelings, they will happily make up the news to reinforce your confirmation bias.

The Royal Society's motto "nulius in verba" is becoming more true by day.

> they will happily make up the news

How are you possibly making that claim? There is nothing to support that.