Entertainment Gab

New work from MIT researchers peers under the hood of an automated fake-news detection system, revealing how machine-learning models catch subtle but consistent differences in the language of factual and false stories. The research also underscores how fake-news detectors should undergo more rigorous testing to be effective for real-world applications.

Popularized as a concept in the United States during the 2016 presidential election, fake news is a form of propaganda created to mislead readers, in order to generate views on websites or steer public opinion.

Almost as quickly as the issue became mainstream, researchers began developing automated fake news detectors — so-called neural networks that “learn” from scores of data to recognize linguistic cues indicative of false articles. Given new articles to assess, these networks can, with fairly high accuracy, separate fact from fiction, in controlled settings.

One issue, however, is the “black box” problem — meaning there’s no telling what linguistic patterns the networks analyze during training. They’re also trained and tested on the same topics, which may limit their potential to generalize to new topics, a necessity for analyzing news across the internet.

In a paper presented at the Conference and Workshop on Neural Information Processing Systems, the researchers tackle both of those issues. They developed a deep-learning model that learns to detect language patterns of fake and real news. Part of their work “cracks open” the black box to find the words and phrases the model captures to make its predictions.

Additionally, they tested their model on a novel topic it didn’t see in training. This approach classifies individual articles based solely on language patterns, which more closely represents a real-world application for news readers. Traditional fake news detectors classify articles based on text combined with source information, such as a Wikipedia page or website.

“In our case, we wanted to understand what was the decision-process of the classifier based only on language, as this can provide insights on what is the language of fake news,” says co-author Xavier Boix, a postdoc in the lab of Eugene McDermott Professor Tomaso Poggio at the Center for Brains, Minds, and Machines (CBMM) in the Department of Brain and Cognitive Sciences (BCS).

“A key issue with machine learning and artificial intelligence is that you get an answer and don’t know why you got that answer,” says graduate student and first author Nicole O’Brien ’17. “Showing these inner workings takes a first step toward understanding the reliability of deep-learning fake-news detectors.”

The model identifies sets of words that tend to appear more frequently in either real or fake news — some perhaps obvious, others much less so. The findings, the researchers say, points to subtle yet consistent differences in fake news — which favors exaggerations and superlatives — and real news, which leans more toward conservative word choices.

“Fake news is a threat for democracy,” Boix says. “In our lab, our objective isn’t just to push science forward, but also to use technologies to help society. … It would be powerful to have tools for users or companies that could provide an assessment of whether news is fake or not.”

The paper’s other co-authors are Sophia Latessa, an undergraduate student in CBMM; and Georgios Evangelopoulos, a researcher in CBMM, the McGovern Institute of Brain Research, and the Laboratory for Computational and Statistical Learning.

Limiting bias

The researchers’ model is a convolutional neural network that trains on a dataset of fake news and real news. For training and testing, the researchers used a popular fake news research dataset, called Kaggle, which contains around 12,000 fake news sample articles from 244 different websites. They also compiled a dataset of real news samples, using more than 2,000 from the New York Times and more than 9,000 from The Guardian.

In training, the model captures the language of an article as “word embeddings,” where words are represented as vectors — basically, arrays of numbers — with words of similar semantic meanings clustered closer together. In doing so, it captures triplets of words as patterns that provide some context — such as, say, a negative comment about a political party. Given a new article, the model scans the text for similar patterns and sends them over a series of layers. A final output layer determines the probability of each pattern: real or fake.

The researchers first trained and tested the model in the traditional way, using the same topics. But they thought this might create an inherent bias in the model, since certain topics are more often the subject of fake or real news. For example, fake news stories are generally more likely to include the words “Trump” and “Clinton.”

“But that’s not what we wanted,” O’Brien says. “That just shows topics that are strongly weighting in fake and real news. … We wanted to find the actual patterns in language that are indicative of those.”

Next, the researchers trained the model on all topics without any mention of the word “Trump,” and tested the model only on samples that had been set aside from the training data and that did contain the word “Trump.” While the traditional approach reached 93-percent accuracy, the second approach reached 87-percent accuracy. This accuracy gap, the researchers say, highlights the importance of using topics held out from the training process, to ensure the model can generalize what it has learned to new topics.

More research needed

To open the black box, the researchers then retraced their steps. Each time the model makes a prediction about a word triplet, a certain part of the model activates, depending on if the triplet is more likely from a real or fake news story. The researchers designed a method to retrace each prediction back to its designated part and then find the exact words that made it activate.    

More research is needed to determine how useful this information is to readers, Boix says. In the future, the model could potentially be combined with, say, automated fact-checkers and other tools to give readers an edge in combating misinformation. After some refining, the model could also be the basis of a browser extension or app that alerts readers to potential fake news language.

“If I just give you an article, and highlight those patterns in the article as you’re reading, you could assess if the article is more or less fake,” he says. “It would be kind of like a warning to say, ‘Hey, maybe there is something strange here.’”

Sega's Yakuza Kiwami 2 is about history, how it shapes our lives and how its cruel absence can leave an all-consuming void that pulls us to our fates. This remake of 2006's Yakuza 2 acknowledges the potential that history has for pain and heartbreak, but it argues that we live fuller lives when we turn to face the things we've put behind us. It doesn't, then, cheapen the intended swan song for longtime protagonist Kazuma Kiryu in Yakuza 6: The Song of Life. By revisiting Kiryu and friends at an earlier point in their story, it fills in the historical blanks for new fans that were brought in by the franchise's resurgence, enriching our understanding of what drives these characters.

In keeping with such themes, it seems apt that looking backward by way of Yakuza Kiwami 2 has strengthened the overall act of playing a Yakuza game. The series, after all, doesn't change much from one entry to the next: Between story beats for a winding gangster melodrama, each game finds Kiryu running around a Japanese city, getting into street fights, playing minigames, and helping bystanders solve problems. But Yakuza 6 was something of a rocky transition for the series, making a lot of concessions in its upgrade to a new game engine. While Kiwami 2 is built on the same engine, there are fewer compromises here. This is a more complete game than its predecessor, bumping up the number of minigames as well as hilarious side stories that contrast Kiryu's stoicism with things like a modeling photoshoot, voice acting, porn tapes, and a fetish for wearing a diaper and being treated like a baby.

Combat, too, is an improvement over that of Yakuza 6. Punches fly faster this time around, and Yakuza Kiwami 2 re-introduces features from earlier games, like a wider moveset and the ability to store weapons. Fights in the Yakuza series are always an over-the-top pleasure, an impossibly dramatic flurry of vicious strikes and thunderous counters from characters who spurt entirely unacknowledged red and blue flames. But to a certain degree, the progression still feels overly slow in an attempt to paper over the repetition of these otherwise satisfying conflicts. Even the most basic moves and the most necessary upgrades are locked behind plot points and level caps.

One of this game's improvements, though, was part of the original Yakuza 2. Despite its humanistic outlook, the series all too often treats women as little more than props to be threatened or beaten, a problem exacerbated by Yakuza 6's failure to let women speak much on the game's ostensible themes of parenthood. Kiwami 2 pairs Kiryu for much of the story with Kaoru Sayama, an Osaka cop infamous for her dogged opposition to organized crime. She's one of the game's most fleshed-out characters, not only backing Kiryu up in fights but embodying the main story's themes as she probes her own history, grappling with whether she might sleep easier knowing nothing at all.

The quality of Sayama's portrayal doesn't, unfortunately, preclude other issues. A well-intentioned side story about a trans woman includes pervasive—and seemingly oblivious—deadnaming in the game's text windows. The gears of the main plot turn on a general suspicion of outsiders that, with its portrayal of the antagonistic Korean Jingweon mafia, edges uncomfortably close to xenophobia. The story simply doesn't afford foreigners the same depths of humanity as the locals, characterizing the Jingweon as single-minded fodder so dedicated to savage violence that they can't be dealt with in the same way as more rational, homegrown yakuza.

The Jingweon's lack of development points to a more widespread issue in the game: Apart from Sayama, few characters or factions here get enough of a spotlight to leave an impression. Only villain Ryuji Goda, a blond yakuza who's also from Osaka, is particularly memorable, and even then, it's more for his charisma and wicked sideburns than anything he's actually given to do. There's an obsession here with plot twists that mostly dumps betrayals and revelations on the shoulders of characters who feel half-formed or outright inconsequential. For as much as Yakuza Kiwami 2 is a better, more confident game than Yakuza 6, the series still has plenty of room to grow.

Federal government employees affected by the government shutdown can get some free entertainment in New York.

The Metropolitan Opera says it is offering free tickets to people with government IDs from Saturday through the end of the month. A pair of tickets per performance can be reserved in advance by telephone or in person at the Met box office.

Twelve performances are available: for Bizet’s Carmen, Mozart’s Don Giovanni, Debussy’s Pelleas et Melisande, Cilea’s Adriana Lecouvreur and a double bill of Tchaikovsky’s Iolanta and Bartok’s Bluebeard’s Castle.

Ugly Betty‘s sisters are going head-to-head: Ana Ortiz will reunite with her TV sibling America Ferrera on the latter’s NBC comedy Superstore.

RELATED STORIES

Per Give Me My Remote, Ortiz will guest-star in a May episode as Claudia Lankow, a no-nonsense senior vice president of operations at Cloud 9. She holds the fate of the Ozark Highlands store, where Ferrera’s Amy is now the manager, in her hands.

Ortiz — who currently co-stars in ABC’s Whiskey Cavalier — is the second Ugly Betty alum to visit the show after Tony Plana, who is set to return later this season as Amy’s father.

Ready for more of today’s newsy nuggets? Well…

* Disney+ is developing a series based on the 1993 film The Sandlot that would be set in 1984 and follow the children of the original characters, our sister site Variety reports. The movie’s co-writer and director David Mickey Evans will pen and executive-produce.

* The ATX Television Festival — taking place June 6-9 in Austin, Texas — will host a Veronica Mars panel for the upcoming Hulu revival. Panelists will be announced at a later date.

* Fox has ordered First Responders Live, an unscripted series from uber producer Dick Wolf that promises “an unprecedented look at fearless first responders as they answer emergency calls across the country.” The series premieres Wednesday, June 12 at 9/8c, with Josh Elliott as host.

Which of today’s TVLine Items pique your interest?

For those who can't make it the second weekend of Coachella 2019, YouTube will be offering a first-ever curated live stream of performances from April 19-21, including Kanye West's highly anticipated Easter Sunday Service.

West's gospel-inspired performance will take place at "The Mountain" inside the festival's campgrounds at 9 a.m. on April 21. "The Mountain" is a hillside inside the Coachella campground area that has become a popular spot to watch sunsets.

YouTube's live stream of Coachella's first weekend started on Friday (April 12) and runs through Sunday. The live stream will pick up again on Friday (April 19). Those interested should subscribe to the channel for updated schedule information.

Coachella organizers first announced West's Sunday Services show on March 31, posting a video on Twitter with the rapper telling the audience at one of his previous services, "We were out in Palm Springs and they took us to a little campground, because we were thinking about a little performance in Palm Springs, just a little one. Then they had a mountain. He had a mountain waiting for us."

YouTube's curated live stream of Coachella's second weekend will be hosted by KCRW's Jason Bentley. It will "take a deep dive into the festival experience and deliver fans a slate of original content," including encore and live performances, artist commentary, mini-docs, and animated adventures, according to YouTube's blog.

On 26 December 2018, the website America’s Last Line of Defense (LLOD) ran this “breaking” news item under the headline “Mike Pence will Resign ‘At The Request of the President’”:

According to our sources in Washington, President Trump will ask Vice President Mike Pence to resign his position as of the 1st of the year. An assistant to an aide to White House Director of Information and Propaganda, Art Tubolls, told our correspondent in the field:

“President Trump says that Pence is the only ‘establishment Republican’ left on his staff, and that the departure of Pence officially drains the swamp, just like he promised. The Executive branch is now full of people the President trusts, not a bunch of crusty old politicians. Mike Pence did a great job, but if we’re going to take the rest of America by force after 2020, we’re gonna need someone in the VP job with a little more gusto. Someone like Donald Jr. or even Ivanka. The Senate is ours. We can break any rule we want and confirm any candidate we want. Why not make the White House all Trump?”

For all that Americans have grown accustomed to the abrupt comings and goings of White House officials under President Trump, however, this announcement rang false even to those who didn’t make it past the headline to read the more ridiculous declarations in the body of the article. For the most part, Mike Pence is seen by the public as the ultimate Trump loyalist (never mind an anonymously sourced report in the New York Times to the effect that the president was privately questioning Pence’s allegiance in November).

As for the rest of the story, it’s true that in the event of the vice president’s resignation Trump could appoint whomever he wants to the position, but it would require Congressional approval — something Trump’s children, should he appoint one of them, would be unlikely to get.

We note, too, that the quoted statement is attributed to Art Tubolls, who is described as an “assistant to an aide to White House Director of Information and Propaganda.” Neither the man nor the position exists. “Art Tubolls” is a fictional character that frequently appears in the pages of LLOD. In another recent article on the site, Tubolls was named as the funeral director handling the interment of the late Penny Marshall.

LLOD is part of a network of websites that publish so-called “satirical” content calculated to incite partisan rage and garner advertising revenue. A disclaimer on the website warns users that everything published there is fictional:

Everything on this website is fiction. It is not a lie and it is not fake news because it is not real. If you believe that it is real, you should have your head examined. Any similarities between this site’s pure fantasy and actual people, places, and events are purely coincidental and all images should be considered altered and satirical. See above if you’re still having an issue with that satire thing.

No such warning appears in social media posts that link to the site’s content.