Headlines
One of the issues in all of the concerns about “fake news” is the way headlines are used on the Internet. Newspapers and magazines certainly took the construction of headlines into account to drive readers into the stories. But on the Internet, headlines have become that and more. They are the links themselves that fly around the Internet and “convert” someone into coming to your site and reading a story. They are “clickbait.” If we want to address the veracity and authenticity of content on the Internet, we might want to start with headlines.
I’ve had my issues with headlines for years. Many years ago, I allowed a number of publications to repost content I write here at AVC on their online publications. The publication that does that most frequently with my content is Business Insider. You can see the hundreds of posts that BI has republished on my author page at Business Insider. When they started doing this maybe seven or eight years ago, I would notice that they would leave my post intact, verbatim, but rewrite the headline. It would drive me crazy because I view the headline as an integral part of my post. I think about the words I use to title my posts. So I would send them angry emails and most of the time they would change it back. But it was a lesson in the difference between a headline that I liked and a headline that would drive clicks.
I also have seen hundreds of stories written about me, USV, and our portfolio companies that have sensational and often inaccurate headlines followed by stories that are essentially correct and well reported. It drives me nuts but I don’t often do much about it.
It makes me think that someone, or some company, or some open source community ought to build software that parses headlines and the stories that follow and rate them for how well the headline represents the article. That “headline veracity ranking” could then be offered to anyone who presents headlines to readers. That would be social media like Facebook, Twitter, Reddit, etc. That would be email applications and browsers. That would be search engines. Etc, etc, etc.
It would be nice to see some competition in this sector so that one company doesn’t become the arbiter of what is an accurate headline and what is not. That doesn’t sound like a good outcome. But if this is done via open source, or is community powered in some way, this could be a very helpful tool in getting publishers to behave and represent their stories accurately.
And that would be a wonderful thing for the Internet.
Comments (Archived):
But would any online publication actually use such a tool? I know New York Times and others use Optimizely to AB test their headlines for clicks.
Yeah I mean i get fred’s frustration, but this is almost like “i wish people would eat vegetables and not ice cream” and then people wouldn’t be fat. Sure, lets change human nature where possible, but lets acknowledge that is what you are asking for. Its not a bug how things currently, its a feature to get more attention. Fred’s trying to fight nature, and nature rarely loses. This would work for a niche audience, no doubt, and that is great, but no way would work on a broad scale
Fred you should change this headline to:-“Fred Wilson destroys Business Insider”-“top 3 ways the worlds most famous VC uses headlines to trick people”-“How to write a headline viewers will click on”-“VCs reveal their secrets to headline optimization”-“Fred Wilson goes to war with mainstream media”-“Fred Wilson basically accuses Henry Blodget of Insider trading in latest headline story”-“AVC pleas with media to end tricking idiots with misleading headlines just to boost CTR to advertisers”-“Wilson to media: ‘stop doing THIS, or else…’ “-
Still, ya gotta love the optimism. It’s Freddy being Freddy.
Touché
We can add AB testing to my list of “Descartes’ binary logic that quantifies but doesn’t help us also qualify things.” https://uploads.disquscdn.c…
only if the distribution system (search, social, etc) penalized them for bad headlines
Are Duck Go Go or Twitter’s AI smart enough to differentiate bad from good headlines and then to penalize? https://uploads.disquscdn.c…If there is no existing data set and no existing NatLang framework to train the machines to differentiate between why+how Content A is bad or good from Content B, none of the AI is smart enough.Google, Microsoft, Facebook and others already open source a lot of parsers.If those parsers were adequate … Google et al wouldn’t have fake news problems and their $ million AI researchers would already have cracked the Natural Language Understanding problem.* https://github.com/tensorfl…
Ah, now that makes sense to me!
For this to happen, everyone (brands, publishers, social platforms) has to stop emphasizing “engagement” which, on a metrics basis, is really just any kind of response at all. And more provocative heds get more responses.
what if engagement was made more specific in terms of response?Granted, I’m not sure if it help because responses could still not have the valence needed for true news, but…
Fred can correct me if I interpreted wrong, but I think he meant that Google, Facebook, Twitter, Reddit et al would use these tools to automatically penalize clickbait titles. It would create a big incentive to write better, more accurate headlines.
I like BI, but only b/c I like to read headlines only, and make snap decisions about what’s probably in the article. And I’m guessing my ADD is exactly what makes BI popular. An endless sea of clickbait that scrapes real content sources, puts key points in summary bullet-point lists, and somehow content creators are OK with it…
Have you seen this yet?https://www.cnet.com/news/f…
“Now you can map the web of lies.A beta version of Hoaxy, a search engine designed to track fake news, was released Wednesday by Indiana University’s Network Science Institute and its Center for Complex Networks and System Research. Hoaxy indexes stories from 132 sites known to produce fake news, such as WashingtonPost.com.co and MSNBC.website, and allows you to see how these sites’ links spread across social media.”
I love everything* about this but especially the name!*except the fact that we need it in the first place, but we’ve always needed some type of veracity mechanism – the only difference now is scale & speed of proliferation
I wish we’d stop calling this ‘fake news’ and start calling it Propaganda because that’s really what it is.
Indeed, it is.
veracity and clickbait language aren’t mutually exclusive.emotive headlines are essentially a choice of style, not necessarily at the expense of accuracy.
Absolutely, but we all know there’s lots of clickbait which is completely deceptive. Emotional headlines can be good, if accurate.
sure, so we could solve the deceptive issue but still have a world full of clickbait headlines as they are factually accurate yet language amped to the nth degree.
Oh, there are lots of complexities. Satire, emotion, kindness. Difficult to parse.
Even if not completely deceptive, they create ‘Anchoring bias’. The headline is the anchor to the article that follows.Paper in the journal of applied psychology, December 2014 – ‘The effects of subtle misinformation in news headlines’.From the abstract – “We demonstrate that misleading headlines affect readers’ memory, their inferential reasoning and behavioral intentions, as well as the impressions people form of faces. On a theoretical level, we argue that these effects arise not only because headlines constrain further information processing, biasing readers toward a specific interpretation, but also because readers struggle to update their memory in order to correct initial misconceptions”.http://psycnet.apa.org/psyc…
lovely
I contribute often to Inc and Entrepreneur. I also enjoyed using my headline as a window into the article, and I often took liberties to make it creative and even entertaining. Most of the time, however, my headline gets changed by the editors. In their defense, they don’t every create a headline that misrepresents the article, but rather leverage the incredible data to understand articles readers react to most. We have even had webinars about it. Personally, I like that they edit the headline, as it inevitably leads to more eyeballs to see the content, which is what I (and the magazines and advertisers) want.
Also, I’d be interested in the treatment of satire which, I’m assuming, might be difficult to assess algorithmically.
As long as the publisher is rewarded for getting a reader to a page by any means necessary they are not incentivized to ensure quality nor veracity.
Or in order to get there first/fast. https://uploads.disquscdn.c…
Imagine if all the other papers of record shared that headline that day? Would we have gotten to the truth and how?
That’s while advertising rules the online world. When subscription becomes valuable alternative quality might become more important. When&might. Internet needs fixing, by people, for people..
Steem is about content creation and payment, but I wonder if its model changes the dynamics of truth and reward?
Yes – I think that would be a more fitting path for them given the economic dynamics of quality content production. I see a high-end tier of subscription-driven content creators re-emerging (like TheInformation, but also not excluding some of the traditional media journalism stalwarts) and readers better appreciating their economic role in support of free, independent reporting. Part 2 of this will be supporting individual journalists (via methods like Patreon) and Part 3 would be rapid organized (crowd)funding of specific investigative journalism projects (once again an area where blockchain can prove pivotal).We’re now in that ugly phase where we realize the limitations of existing methods of monetization but haven’t quite hotwired their next evolution yet.
subscription doesn’t equal reading/engagement though.
It doesn’t, but it certainly incentivizes greater independence of what’s covered as well as quality befitting the target audience.Noone & nothing can make people read & learn what they should, sadly.
yeah, but if they are penalized by “distribution” like search, social, or something else, then maybe they would change their tune
Who can penalize them? I think only Google could
Facebook, Twitter, Reddit and other internet watering holes could also do multiple actions:1. indicate a clickbait title with a small visualization next to submission (reddit actually provides mechanism to do this kind of things manually)2. provide users an option to filter out clickbait submissions 3. automatically penalize sources that use clickbait headlines
Google also powers the advertising stacks of most major global publishers so their incentive to fix this is somewhat mixed at best. However, it’s an opportunity for Bing who is free(er) of such potential conflicts of interest and folks like DuckDuckGo.
True, but these other 2 have little impact because their traffic is tiny, compared to Google.
Bing is about 15% and I assume DDG is ~1%. I don’t know many successful businesses that would ignore 15% of the market.And, let’s not forget, Google was a late-comer to the search space. The difference was that they did it so much better than everyone else. If another search engine produces better results that Google, I think people will switch. After all, there really isn’t much lock-in to using Google for search.
The trouble is that all of those distribution channels are also incentivized for engagement (whichever definition of engagement they individually adhere to): Google powers the advertising layer of most publishers, Facebook needs content so users would keep logging in, and the something else used to be the strength of a media brand’s reputation that’s now been almost universally diluted.We’re in between two business models: the old, ad supported one and a new one that once again returns to audience-building. The trouble is how to swap out all 4 tyres of a car that’s moving down the highway.
You can’tAnd historically, yellow journalism was the journalism that made money. It might be that a stable ride was a brief moment in history.
Completely agree Ana.Like him or hate him Ryan Holiday was behind those epically successful American Apparel ads. He wrote an excellent book “Trust Me I’m Lying: Confessions of a Media Manipulator”. In it he argues the economics of the web is returning us to an era of yellow journalism.To Fred’s point, how accurately a title reflects the content of an article is really a matter of opinion. Of which there would be a spectrum. Since it’s not true or false, you cannot get software to do this.This is why we have lawyers. Language can be interpreted in a myriad of ways. It does not nor cannot frame “truth”.
I reviewed Holiday’s book a while back on my blog:http://www.iheavy.com/2014/…
Yes, I wrote about misaligned incentives & media’s innovators dilemma in-depth here: https://www.baekdal.com/ins…More so than before we are now primed for the era of the media manipulator: in everything from the highest levels of politics to more mundane things like teatox “influencers” on social media. I’d argue it’s not the economics of the web but rather the economics of attention – attention is finite & time-constrained, we only have so much of it during any given day. By making it currency we incentivize behaviors that are meant to capture our attention even if it’s for a few seconds by any means necessary. Any means necessary readily translates to more and more extreme shock tactics.
At what point do publishers begin to notice a significant number of potential readers have stopped clicking on clickbait? Is there a way to measure non-clicks by clickbait-hating readers? Is there a way of measuring the value gained by clickbait-loving readers versus the value lost via the exodus of clickbait-hating readers? And there has to be some sort of learning-curve that every reader sits on, whereby today’s clickbait-lover becomes tomorrow’s clickbait-hater. I discovered this article not through clickbait-y headlines but through highlights by readers on highly.co, which acts as a content-centric way of discovery.
I wrote about some of that, along w/ an analysis of the publisher innovation dilemma & misaligned incentives here: https://www.baekdal.com/ins…
You’re seeing this from one side, your side.First, I prefer to use the term Title, not Headlines.Bona fide (reputable) publishers have had a knack for picking good titles that strike a balance between keeping the reader interested and reflecting a linkage to the content bring served. When I was a columnist for Business 2.0 in ’97-00, I would let them write or rewrite my articles titles, and they did a better job than me, because that’s what good editors do.In your case, you have a daily loyal readership and pretty good SEO, which means that AVC is not so dependent on tinkering with posts titles.That said, I agree with the spirit of your post, which is to say that generally, on the Internet, titles have become sensationalized. I think we can thank the Drudge Report and the Huffington Post for pioneering the link bait tactics, later practiced by the Buzzfeed, BI & other pubs that emphasize the headline more than the substance of the content.But this “rate them for how well the headline represents the article” would be a noble goal. However, I’m not sure that it would force publishers to behave or change their tactics. How they pick their headlines is directly related to the kind of content they publish. If you forced BI or HuffPo to change their headlines style, 90% of their content wouldn’t be clicked on or read, and they might go bankrupt.Instead of rating each article on that “authenticity” scale, I think there is merit in just rating the publication as a whole based on the sum of their practices, because they are consistent with it. For eg, I know that the NY Times titles will tilt on the boring scale, and you have to force yourself to click them, and there is good stuff underneath. And I know that for other pubs, the headlines bait detector needs to be elevated, and it is what it is.
What is a bonafied publisher?You mean you and I of course and 1 gazillion others?
I meant – “reputable”, as a better choice of words i should have used. Again, I use the NY Times as a good example of a reputable publisher who strikes a good balance where the headlines accurately reflect the content (although tilting on conservative).
of course I agree william but does this really work any longer?the majority of stuff you the master curator scrape out every day is from blogs not the traditional publishers.if fact while the publishers are certainly the purveyors of true journalism at their best, news belongs to the rest of us.
Yup. I make mental notes of those that sensentionalize their titles, and that ups my bullshit detector on the next article. Most people are quite predictable in their originality styles, authenticity levels and degrees of insights.
Yup !Most readers over time will evolve better BS antennae. I suspect in the long run that user evolution is the only organic solution up to the task at hand ?
Yup, like a restaurant, if they serve you bad food or give you bad service, you’ll think twice about returning.
Hmmn … could the devs I’ve befriended in SF and I do a crowd-sourced “veracity ranking” plugin that every publisher can use? Well …Q4 2013, I met Mark Thompson (former DG of the BBC, now CEO of NYT) at a media conference in NYC. I gave a 10-second demo of my system.His immediate response was: “Wow, that’s super-cool!”I asked him, “How would NYT like to be able to measure readers’ perceptions of each article, comment, headline so they could tailor content to users’ preferences better?”He invited me in to meet their User Insights team. Bear in mind that NYT was re-organizing its Digital Strategy after Thompson was hired.The meeting went well but was followed by them deciding to stay within their comfort zones.Dec 2016, this is how NYT surveys to try to understand its readers and their preferences (aside from obvious clickbait metrics):https://uploads.disquscdn.c…https://uploads.disquscdn.c… https://uploads.disquscdn.c…https://uploads.disquscdn.c… https://uploads.disquscdn.c…Meanwhile, in Q4 2012 I met RCS Mediagroup of Italy who publish Corriere della Sera.Yes, I quoted Da Vinci back to his own countrymen who couldn’t see or connect his work with the Digital Age and its metrics. Lol.https://uploads.disquscdn.c…After meeting me, they changed how they profile readers and the way they METRIFY EVERY ARTICLE. https://uploads.disquscdn.c…
Presume the “veracity ranking” you reference is a perceptual rating, rather than a fact-based content authentication metric, which is difficult to track and saddled w/ subjectivity and nuances. That said, if you can combine and weight a site’s traditional quant metrics (e.g., u/v’s, time spent, page views) w/ a range of qualitative measures (e.g., authentic, believable, credibile, trustworthy, truthful, reputable), then you’ve potentially got something that’s far more discriminating for advertisers than current media planning tools.For example, on a weighted (if not absolute) basis, the NYT, WaPo, The Guardian, etc., would potentially dwarf HuffPo, BI, NYP and/or other pubs that frequently engage in clickbait.This is something my former colleague Jason Kint at Digital Content Next should be spearheading.http://digitalcontentnext.o…
Thanks, I have invented a system to measure qualia and quanta.You mean this Jason Kint: https://uploads.disquscdn.c…* businessinsider.com/dcn-let…So here are the HUGE bottlenecks that the orthodoxy of about 450+ years of Descartes and Bayes’ methods has created wrt our and the machines’ ability to understand content.https://uploads.disquscdn.c…What’s fact and what’s fake are both subjective and relative, and the tools of Cartesian logic and Bayesian probability distributions and percentages can’t enable the machines to tell which is which.https://uploads.disquscdn.c…http://qz.com/843110/can-ar…In case Jason Kint isn’t aware, Silicon Valley’s best and brightest have no idea how to solve the Natural Language Understanding problem:* https://www.technologyrevie…Specifically wrt whether Google’s Deep Learning can get the machines to understand us: https://uploads.disquscdn.c…* https://www.linkedin.com/pu…Wrt whether Facebook’s top engineers can solve the fake news problem with their version of a veracity ranking metric:https://uploads.disquscdn.c…@fredwilson:disqus ‘s ask of “software that parses headlines and the stories that follow and (the crowd) rates them for how well the headline represents the article” is what FB is experimenting with.Except that that rating scale (that’s been a staple of market research, advertising and political polling since 1932) is deeply flawed.Moreover, except that the various concept semantic graphs built up by the major techcos are also not enabling the machines to understand meanings in content for coherent parsing.Having pointed out some of the reasons Google & Facebook’s top AI people are stuck on how to solve the “fake news” problem …Sure, it’s possible to change Advertising and Media Buying to systems balancing qualification with quantification instead of pure quant clickbait and gaming probability / power law curves.https://uploads.disquscdn.c…
Can I get in touch with you about this offline, thanks.Ray Kurzweil of Google on AI, July 2016: “Emotions and human language are the most cutting-edge of human intelligence. Emotions are not a sideshow. Consciousness is subjective. And, fundamentally, that (subjectivity) is not a scientific concept.”All data is objective+subjective and, frankly, the big techcos don’t know where to start to build the appropriate subjective measurement tools (the qualifiers that can complement the standard quant techniques that already exist).My email is twainventures[@]gmail.
The less friction to collect data, the better.
Thanks, I baked that into my system.Here’s the paradox of when investors give the message of less friction. Designers and devs then go straight to, “Well, that means 1-click, more engagement, hockey-stick. Investors will love that.”But that’s how users end up with a like button, a swipe right, 5 emojis or a sliding scale from strongly disagree to strongly agree.And that UX sucks us all into the abyss that is clickbait.See? So investors, designers, developers and users alike have played a part in creating the clickbait culture.To change from that culture, all the agents involved would have to change mindset and behaviors.
“Instead of rating each article on that “authenticity” scale, I think there is merit in just rating the publication as a whole based on the sum of their practices, because they are consistent with it.”Agree. Ever thought about why Google removed and disabled the PageRank system? PageRank was the best system to rate quality of web pages and blog sites. Now it’s a mess out there. But PPC revenue is still growing. As long as Clicks is the payment reward, it will be more mess. Someone should create a new quality reward system that replace clicks as a business model. Facebook/Google fair traditional media will get strong again. I think “click reward” will be challenged over the next decade.
Yes, agree w your comment too. Google controls lots here.
They haven’t fully gotten rid of pagerank.And it was debatable if pagerank was great. Remember content farms?
You are right. PageRank is/was not perfect and there will always be need for manual curating somehow. Is content farms still a problem? Cheating should not be rewarded (great content should). However people decide what they want, so market adjust.
Wonderful idea!I’d take this a step further, though, and say someone should build a browser plugin that looks at each story, parses it, develops a new headline, and replaces the original one. This would be like putting on goggles that make clickbait disappear from your life.Consumers should have a right to control their own experiences on the web. They should be in control of their browsers and in control of how content is displayed on all of the sites they visit.
i love that idea
Except then it could change _your_ headlines, too.
shock blogging is nothing new.difference is it used to be the domain of the tabloids, now it is simply part of the fight for attention.I agree though a mess but honestly with current focus on page views i don’t see this changing anytime soon.
community/open source would be great. if news is to keep the public informed we shouldn’t count on corporations to end the shenanigans, way better for them that people are entertained or pumped up in some way. “real news” seems pretty broken too.
I would like to see credibility rating for publishers on the internet. The rating should be clear and easy to see placed next to the headline. This will deter publishers from posing “FAKE NEWS”.
See @savedyouaclick on Twitterhttps://twitter.com/SavedYo…
what’s a headline for? there’s no content bundling here. every morning there’s a content choice of just one post. signposting is not needed. compel people to read for comprehension.experiment. be unconventional. write no headline. observe the response. await feedback. what’s to lose?
There is a company/small group of people that recognizes this issue too and tries to present the news in a much more objective, data based way. https://www.ethicalmedia.org.Full disclosure – I know because I used to write for them. They are still in early stages but their business model is different (subscription based) and does not rely on advertising model or clicks so they are able to focus on quality. One of the things they do is analyze headlines, comparing various headlines of a story and comments on the issues with them.
As long as qualitative criteria in the ad buying process continues to go south (or ignored) this will never be fixed. Today it’s all about eyeballs and engagement, authenticity, embellishment and accuracy be damned. Agencies and/or industry trade orgs have the power to fix or at the very least control this, if desired. Qualitative metrics need to be elevated to a level comparable to other buying criteria.
THIS.
Sounds fairly straightforward for a neural network. How large of a market do you think exists for this (especially given the “click-bait” race-to-bottom content market)?
maybe no market. maybe it has to be something that is done open source
you can’t really run a neural network without open source data as well.
I think “accurate” is the loaded term here. I will call out the NY Times since I’ve noticed them being the worst purveyor of dual headlines – one for the click and one for the history books. The former is often a call out from a comment deep in the article (“Trump’s Empire is a Mom-and-Pop Business”) and the other is the factual (albeit longer) “Inside the Trump Organization, the Company That Has Run Trump’s Big World” that is more clearly about the global operation.Headline “optimization” is disconcerting, as it continues the trend toward news as a means of marketing for page views vs. news as journalism for informing the citizenry. The, ahem, government should fund (through Center for Public Broadcasting) or MPC a way to just do a match of headlines clicked to the actually headline within the story. That analysis alone would be eye opening.
i agree
The clickbait headline fix could come from an evolution of business model. In the 90s we talked about hits. Then we paid for impressions and later Google started to charge for clicks. How can we reward great content online without a paywall? Then maybe it will be less tempting to create clickbait headlines, that is noise. One company I advice, we changed focus from online visitors to actually online leads. The management totally changed how to think about content and quality. Profit also increased a lot. A big bet is to beat the click as payment model with a higher quality measured business model.
So what would be the outcome of that “headline/content alinement” process ?1- Good content adjusts headlines to be less sensational.or2- Posted content is adjusted to be more sensational to support traffic diving click-bait links.
NEWS FLASH !Fake news is a black-ops attempt by mainstream media outlets to drive readers to pay walls.( Thats what I heard)
goes back to Simon Sinek and how you make decisions. You make them with your limbic brain, not logic. Hence, the over wrought emotional headlines we see.
Right and where do 99% of data, economics and AI systems start?https://uploads.disquscdn.c…So this is why we shouldn’t be surprised that AI’s incapable of understanding us and our natural language when it’s been designed as the OPPOSITE OF HUMAN INTELLIGENCE.
veracity.ai is available as a URL.https://domainr.com/veracit…Techmeme strikes me as a firm in a good position to sponsor/promote such an effort, as they are re-writing headlines to make them less clickbait-y.Narrative Science has an AI platform called ‘Quill’ that auto-generates prose, and Quill will become a DIY platform some time in 2017 — someone using that platform could create a ‘Seal of Good Housekeeping’ certification for good headline writing (like TrustE).And then imagine Bitly (or similar platforms) aggregating click data that shows the Veracity of headlines — kind of a ‘PageRank for Headlines’ ranking system.And (der, obv) all this gets logged on The Blockchain.$250k ask at a $750k post-money for the MVP?
i am more interested in this being done as some community powered open source sort of thing
‘If I were a younger man…’ <black gloved=”” fist=”” in=”” the=”” air=””>I think you’re seeking an atomic-level version Genius.com (see: A16Z), where snippet-level aggregation becomes the (for-profit) Wikipedia of Veracity. I’m picking up what you’re laying down, and I can see the pieces coming together, and nothing I suggest above is meant to read all open-source-y (sorry if that was unclear).Second Try: $250k for a post-money of $749k?
Crap. I see. You *want* an open-source-y kind of thing. That approach is also open to the above terms (eesh).
The prize to solve the fake news problem (40+ AI researchers with repo on Github) is currently $2000 so USV should chip in $3000 and make it the same prize amount as a typical w/e hackathon:* https://www.wired.com/2016/…* http://www.fakenewschalleng…
It’s a PBC play.There has to be an incentivization+deincentivization model. William makes a good point: “Who can penalize the publishers? Only Google.”Them, FB and the ad exchanges which, as Salt Shaker points out, are not tooled for qualitative parsing but are tooled for “race to the bottom” quant counts of clickbait.I know Delip Rao, the ex-Amazon Echo AI researcher mentioned in ‘Wired’ article on fake news challenge created by Carnegie Mellon professor.The output criteria they’ve set up for the parsing test:1. Boolean fakeness indicator2. Confidence score3. Provenance URL to support4. A confidence threshold for accepting/rejectingIs an example of the type of predictive approach you railed against in ‘New Economics’ post, yet it’s PERVASIVE in tech UX design and Natural Language Processing.It’s one of the reasons the algos won’t be able to parse your headlines and understand your meanings and intents (and respect them) any day soon.I’ll be in SF in Jan for AI conferences.So I’ll catch up with Delip and give the organizers a steer on setting parse test criteria that go beyond Boolean (0-1) and confidence score.Bayes deductive logic (0 or 1) and probability bell curve (confidence score) are the tools NLP has used for parsing headlines (and all content) since AI was born as a discipline in 1950s.If we can’t break their stranglehold on how the best AI minds in the world code and set criteria for content parsing ….We can all look forward to more “fake news.”Fundamental paradigm shifts are needed across maths, tech-AI, economics and data to do something as simple as have a parser that understands your meaning rather than that approximates or pattern recognises it by probability and Boolean logic.
When we were at Circa, we tried to buck the trend on headlines. It comes down to whether the headlines were informative or bait. I still think there is a value that can be provided here – especially as push notifications to a mobile screen can take up more space and contain more of the story today than ever before.http://blog.digidave.org/20…
This headline problem is one of the animating forces for Highly, a highlighter for the internet. Because most people are reticent to leave their social feeds to even visit the full stories (lots of reasons for this), the likebait nature of today’s headlines has more than a misleading effect. It also ensures that social feeds — where we spend the most time — are full of some of the least useful, least representative aspects of the stories they purport to represent.A stream of Highlights instead of Headlines would much better represent its stories. (So we’re building the platform to support it: http://www.highly.co) We articulate the problem and solution a bit more here in this post headlined Fake Reading Opened the Door for Fake News: https://shift.newco.co/fake…Returning to Fred’s riff: how might a community-powered “headline veracity ranking” work? Comparing readers’ top highlight(s) with the headline itself could quickly indicate whether there’s a mismatch between what the pub and the people believe is important. A solid start.
What type of copyright license do you operate under? Would seem, if desired, you could control any undesired re-work.
This headline is very inaccurate (and really not helpful)BitTorrent Live’s “cable-killer” P2P video app finally hits iOShttps://techcrunch.com/2016…But honestly what is worse is this whole Verge article that suggests we are not using P2P to stream Live. This has been reblogged hundreds of times by other sites. We sent them a response but they have not updated the story.http://www.theverge.com/201…Tech media is getting really sloppy.
I have a similar story, with an editor changing my headline.I wrote a Medium piece about discrimination in the workplace. A couple thousand people read it.Fast Co picked it up. The editor wanted to change the headline (with my consent), and I resisted at first, but I relented, suggested some alts, and we’d landed on something closer to what Fast Co wanted. I figured they knew better. Tens of thousands more people saw it, spreading what I think is a worthy message that much further. So, I guess they did know better. Or did they? Who knows, really?The difference is that my original headline (“The Diversity Shield”) was somewhat neutral, as I wanted it to be about the broader issue of discrimination in workplaces in general, and just use my past employer as an example to illustrate. However, Fast Co’s headline felt more like a direct attack (“Company’s Can’t Hide Behind #BlackLivesMatter”), followed by a byline that called my former employer out by name, and added a hint of drama. All the content after the headline was the same but, in an effort to present as unbiased of an argument as I possibly could, I really didn’t want to use any pumped up attack language (the plain facts were damning enough).But who am I to snub my nose at my message being amplified several orders of magnitude? The feed back from people who saw the post on both platforms was equally positive.Should you care that Business Insider changes your headline, if your message is getting amplified? You still benefit, right?Still, I see your point.I’m not making a counterpoint here. I don’t think I even have one to make; I can’t stand clickbait culture, but I also can’t really denounce the system that makes clickbait a thing when I positively benefited from it. But I also think that the intentions of the editors at Fast Co were way, way, way above and beyond the fake news culture that we have, so there’s that, too.I guess its a Catch 22 of sorts.
“USV Expected To Fund Companies With Headline-Improvement Software”
There are a lot of reasons why this happens–some are systemic, some have to do with the individual publishers. But here are a few:Mobile design for news sites is constructed around headline-based discovery mechanisms. It’s rare that you see the lede of a story exposed on a mobile browser. You’re lucky if you get a dek.Publishers are under-resourced so they often use headlines on stories that would also work on Twitter and Facebook because they feel they don’t have the bandwidth to customize for each platform. So you end up with social-friendly but not very sophisticated headlines.A/B testing on headlines is much more commonplace, and not surprisingly, catchier/baitier heds perform better. For the writers and reporters who do the work: if a catchy hed gets you to actually read a story, they’d generally prefer that to you not reading at all because the headline was less sexy.And lastly, I think publishers are having a hard time wrapping their heads around the fact that people often share stories based entirely on the headline–without reading a word of the story. And users often make inferences from the headlines that may have nothing to do with the actual story itself. Journalists tend to think of the story itself as the most important element and the packaging of the story as very secondary, so they assume everyone else does too. And for everyone else, the headline is often enough.
We have an API for this at Agolo, would love to demo it for you!
The current efforts on the distribution side to improve relevancy tend to be broader and non-obvious, such that they only remove the worst-of-the-worst, rather than promote the best.In addition, many distributors benefit from driving higher click-throughs and incentivize this poor behavior with their business models.This is why the publisher’s brand is so important. As a user, you can pick the brand which does not over-promise and under-deliver by continuing to read their content even if it doesn’t promise you the sun.
There are open source partial solutions available/papers with sample code around.Very factual isn’t moneymaking for anyone though because of the way traffic moves.:/
Indeed, if we separate the bait from the content, the content can stay accurate, and readers who feel misled when they click on the bait will be less likely to click on that source in future, giving the source some incentive to stop being misleading, though it might take a while. I’m thinking that Fred should not permit anyone to repost his content unless they agree not to alter the headline without permission. That won’t help with the second situation he mentioned, where the other site has a good article with a bad headline, but all we can hope for there is that readers will ultimately prefer sites with accurate headlines.
Hey @jhabdas:disqus , thanks for sharing the headline analyzer! And @RohanSJ:disqus , you’re right: It’s not click bait if your content delivers on the promise your headline makes. Love it. -Nathan from CoSchedule