Facebooks redoubled AI efforts wont stop the spread of harmful content – VentureBeat
Posted By admin on November 13, 2020
Facebook says its using AI to prioritize potentially problematic posts for human moderators to review as it works to more quickly remove content that violates its community guidelines. The social media giant previously leveraged machine learning models to proactively take down low-priority content and left high-priority content reported by users to human reviewers. But Facebook claims it now combines content identified by users and models into a single collection before filtering, ranking, deduplicating, and handing it off to thousands of moderators, many of whom are contract employees.
Facebooks continued investment in moderation comes as reports suggest the company is failing to stem the spread of misinformation, disinformation, and hate speech on its platform. Reuters recently found over three dozen pages and groups that featured discriminatory language about Rohingya refugees and undocumented migrants. In January, Seattle University associate professor Caitlin Carlson published results from an experiment in which she and a colleague collected more than 300 posts that appeared to violate Facebooks hate speech rules and reported them via the services tools. According to the report, only about half of the posts were ultimately removed. More recently, civil rights groups including the Anti-Defamation League, the National Association for the Advancement of Colored People, and Color of Change claimed that Facebook fails to enforce its hate speech policies. The groups organized an advertising boycott in which over 1,000 companies reduced spending on social media advertising for a month.
Facebook says its AI systems now give potentially objectionable content thats being shared quickly on Facebook, Instagram, Facebook Messenger, and other Facebook properties greater weight than content with few shares or views. Messages, photos, and videos relating to real-world harm, like suicide, self-harm, terrorism, and child exploitation, are prioritized over other categories (like spam) as theyre reported or detected. Beyond this, posts containing signals similar to content that previously violated Facebooks policies are more likely to reach the top of the moderation queue.
Using a technique called whole post integrity embeddings, or WPIE, Facebooks systems ingest deluges of information, including images, videos, text titles and bodies, comments, text in images from optical character recognition, transcribed text from audio recordings, user profiles, interactions between users, external context from the web, and knowledge base information. A representation learning stage enables the systems to automatically discover representations needed to detect commonalities in harmful content from the data. Then fusion models combine the representations to create millions of content representations, or embeddings, which are used to train supervised multitask learning and self-supervised learning models that flag content for each category of violations.
One of these models is XLM-R, a natural language understanding algorithm Facebook is also using to match people in need through its Community Hub. Facebook says that XLM-R, which was trained on 2.5 terabytes of webpages and can perform translations between roughly 100 different human languages, allows its content moderation systems to learn across dialects so that every new human review of a violation makes our system[s] better globally instead of just in the reviewers language. (Facebook currently has about 15,000 content reviewers who speak over 50 languages combined.)
Its important to note that all content violations still receive some substantial human review were using our system[s] to better prioritize content, Facebook product manager Ryan Barnes told members of the press on Thursday. We expect to use more automation when violating content is less severe, especially if the content isnt viral, or being quickly shared by a large number of people [on Facebook platforms].
Across many of its divisions, Facebook has for years been moving broadly toward self-supervised learning, in which unlabeled data is used in conjunction with small amounts of labeled data to produce an improvement in learning accuracy. Facebook claims its deep entity classification (DEC) machine learning framework was responsible for a 20% reduction in abusive accounts on the platform in the two years since it was deployed and that its SybilEdge system can detect fake accounts less than a week old with fewer than 20 friend requests. In a separate experiment, Facebook researchers say they were able to train a language understanding model that made moreprecise predictions with just 80 hours of data compared with 12,000 hours of manually labeled data.
For virility prediction, Facebook relies on a supervised machine learning model that looks at past examples of posts and the number of views they racked up over time. Rather than analyzing the view history in isolation, the model takes into account things like trends and privacy settings on the post (i.e., whether it was only viewable by friends).
Virility prediction aside, Facebook asserts that this embrace of self-supervised techniques along with automatic content prioritization has allowed it to address harmful content faster while letting human review teams spend more time on complex decisions, like those involving bullying and harassment. Among other metrics, the company points to its Community Standards Enforcement Report, which covered April 2020 through June 2020 and showed that the companys AI detected 95% of hate speech taken down in Q2 2020. However, its unclear the extent to which thats true.
Facebook admitted that much of the content flagged in the Wall Street Journal report would have been given low priority for review because it had less potential to go viral. Facebook failed to remove pages and accounts belonging to those who coordinated what resulted in deadly shootings in Kenosha, Wisconsin at the end of August, according to a lawsuit. Nonprofit activism group Avaaz found that misleading content generated an estimated 3.8 billion views on Facebook over the past year, with the spread of medical disinformation (particularly about COVID-19) outstripping that of information from trustworthy sources. And Facebook users in Papua New Guinea say the company has been slow or failed to remove child abuse content, with ABC Science identifying a naked image of a young girl on a page with over 6,000 followers.
Theres a limit to what AI can accomplish, particularly with respect to content like memes and sophisticated deepfakes. The top-performing model of over 35,000 from more than 2,000 participants in Facebooks Deepfake Detection Challenge achieved only 82.56% accuracy against a public dataset of 100,000 videos created for the task. When Facebook launched the Hateful Memes dataset, a benchmark made to assess the performance of models for removing hate speech, the most accurate algorithm Visual BERT COCO achieved 64.7% accuracy, while humans demonstrated 85% accuracy on the dataset. And a New York University study published in July estimated that Facebooks AI systems make about 300,000 content moderation mistakes per day.
Potential bias and other shortcomings in Facebooks AI models and datasets threaten to further complicate matters. A recent NBC investigation revealed that on Instagram in the U.S. last year, Black users were about 50% more likely to have their accounts disabled by automated moderation systems than those whose activity indicated they were white. And when Facebook had to send content moderators home and rely more on AI during quarantine, CEO Mark Zuckerberg said mistakes were inevitable because the system often fails to understand context.
Technological challenges aside, groups have blamed Facebooks inconsistent, unclear, and in some cases controversial content moderation policies for stumbles in taking down abusive posts. According to the Wall Street Journal, Facebook often fails to handle user reports swiftly and enforce its own rules, allowing material including depictions and praise of grisly violence to stand, perhaps because many of its moderators are physically distant and dont recognize the gravity of the content theyre reviewing. In one instance, 100 Facebook groups affiliated with QAnon, a conspiracy labeled by the FBI a domestic terrorist threat, grew at a combined pace of over 13,600 new followers a week this summer, according to a New York Times database.
In response to pressure, Facebook implemented rules this summer and fall aimed at tamping down on viral content that violates standards. Members and administrators belonging to groups removed for running afoul of its policies are temporarily unable to create any new groups. Facebook no longer includes any health-related groups in its recommendations, and QAnon is banned across all of the companys platforms. Facebook is applying labels to but not removing politicians posts that break its rules. And the Facebook Oversight Board, an external group that will make decisions and influence precedents about what kind of content should and shouldnt be allowed on Facebooks platform, began reviewing content moderation cases in October.
Facebook has also adopted an ad hoc approach to hate speech moderation to meet political realities in certain regions around the world. The companys hate speech rules are stricter in Germany than in the U.S. In Singapore, Facebook agreed to append a correction notice to news stories deemed false by the government. And in Vietnam, Facebook said it would restrict access to dissident content deemed illegal in exchange for the government ending its practice of disrupting the companys local servers.
Meanwhile, problematic posts continue to slip through Facebooks filters. In one Facebook group that was created this past week and rapidly grew to nearly 400,000 people, members calling for a nationwide recount of the 2020 U.S. presidential election swapped unfounded accusations about alleged election fraud and state vote counts every few seconds.
The system is about marrying AI and human reviewers to make less total mistakes, Facebooks Chris Parlow, part of the companys moderator engineering team, said during the briefing. The AI is never going to be perfect.
Continued here:
Facebooks redoubled AI efforts wont stop the spread of harmful content - VentureBeat
- Sacramento GOP call for resignation of member with ties to Proud Boys - ABC10.com KXTV - January 26th, 2021
- Symbols of white supremacy confront Redmond teen - OPB News - January 26th, 2021
- Story Jan. 22, 2021 Faculty voice: Gaming and toxicity - MSUToday - MSUToday - January 26th, 2021
- Jewish communities in New York City and across the country tighten security protocols as threats mount - SecurityInfoWatch - January 24th, 2021
- FBI and Huntsville Police Announce Reward and Billboard Campaign to Assist in Identifying Person of Interest in Vandalism of Jewish Synagogues FBI -... - January 24th, 2021
- Woodland Park man arrested for US Capitol insurrection - FOX21News.com - January 24th, 2021
- FBI and Huntsville police announce reward in Identifying person of interest in vandalism of Jewish Synagogues - WVTM13 - January 24th, 2021
- Parents outraged over students Confederate flag face mask demand action - WKRN News 2 - January 24th, 2021
- QAnon after Trump: Believers of the conspiracy theory at risk of further radicalization, experts warn - CTV News - January 24th, 2021
- Decoding the extremist symbols and groups at the Capitol Hill insurrection - ABC17News.com - January 12th, 2021
- Lawmakers Fear More Violence Ahead Of Inauguration Day : Insurrection At The Capitol: Live Updates - NPR - January 12th, 2021
- The Big Question: Can the U.S. Defuse Violent Right-Wing Extremism? - BloombergQuint - January 10th, 2021
- Opinion: We have to work together to protect democracy from threats - The Detroit News - January 10th, 2021
- Sacha Baron Cohen on Facebook, Twitter and Trump - Variety - January 10th, 2021
- Extremists intensify calls for violence ahead of Inauguration Day - WDJT - January 10th, 2021
- Trump Helped Take Extremist Views From The Fringes Of Society To A Mob Attacking The Capitol - FiveThirtyEight - January 10th, 2021
- US Capitol: Q-Anon, Confederate flag man, and Baked Alaska - here are the people who stormed the building - Sky News - January 10th, 2021
- Domestic terrorism and hate exploded in 2020. Here's what the Biden administration must do. - ABC News - January 1st, 2021
- Arizona hate crimes bounced back in 2019, experts fear a surge in 2020 - Cronkite News - January 1st, 2021
- Weeks-Old Statue of Breonna Taylor Is Battered in Oakland, Calif. - The New York Times - January 1st, 2021
- Hanover police need help identifying possible menorah vandals - The Union Leader - December 27th, 2020
- Boycott urged on firms that assist Israel's occupation - Triple Pundit - December 27th, 2020
- After Dartmouth menorah was vandalized, Hanover police seek more information - Concord Monitor - December 22nd, 2020
- New York bans display of Confederate flag and other hate symbols on state grounds - WDJT - December 22nd, 2020
- 'We're against everything they stand for': LGBTQ-owned clothing company Verillas pushes back after Proud Boys wear its kilts - USA TODAY - December 22nd, 2020
- Thomas O'Brien on Boston's Newest Luxury High-Rise Residential Tower, Bulfinch Crossing Development, Pandemic and 2021 - Boston Real Estate Times - December 22nd, 2020
- Hate groups receive millions of dollars in federal funds - WVTM13 - December 14th, 2020
- 'Is this what we're becoming?': Anne Frank memorial in Idaho, the only one in US, defaced with swastika stickers - USA TODAY - December 14th, 2020
- Opposition Mounts Against Proposed Appointment of Far-right Chairman to Israeli Holocaust Museum - Hyperallergic - December 14th, 2020
- Only yesterday - The Hudson Reporter - December 14th, 2020
- Chelsea FC launches exhibition about Jewish Athletes and the Holocaust - Chelsea FC - December 14th, 2020
- Evergreen Mill Elem. student named 'Kid of the Year' finalist - Loudoun Times-Mirror - December 14th, 2020
- Conservatives flocked to Parler after the election. But its explosive growth is over - KCTV Kansas City - December 14th, 2020
- Anti-Defamation League to honor health secretary Levine ... - December 5th, 2020
- The Anti-Defamation League Is Not What It Seems | Boston ... - December 5th, 2020
- Travel Fairness Now Hosts Webinar with Health Care, Consumer and Travel Experts to Improve Traveler Access to Vital Information During Covid-19 and a... - December 5th, 2020
- ADL and Aspen Institute Announce Two Civil Society Fellowship Classes in Response to the Heightened Need for Civil Discourse in a Divided America -... - November 17th, 2020
- Tens of thousands rally in DC to support outgoing President Trump; at least 20 arrested as protesters clash with counterprotesters - USA TODAY - November 17th, 2020
- White supremacist jailed for 2 years over plot to bomb Las Vegas synagogue, ADL - The Times of Israel - November 15th, 2020
- ADL webinar addresses election, extremes - Cleveland Jewish News - November 13th, 2020
- Jewish orgs. react to presumptive win of Dem. nominee Joe Biden - The Jerusalem Post - November 13th, 2020
- Sharon L. Klein Named to Working Mother's Top Wealth Advisor Moms 2020 - PRNewswire - November 13th, 2020
- Proceed with eyes open - Isthmus - November 13th, 2020
- Tip lines allow voters to report intimidation, extremism at the polls - Chicago Sun-Times - November 3rd, 2020
- Hate Speech on the Rise - PBS39.org - November 3rd, 2020
- 100-year-old Jewish cemetery in Michigan vandalized with red paint - New York Post - November 3rd, 2020
- Who are the 'Proud Boys'? Photos of Austin protesters circulating online - KIIITV.com - November 3rd, 2020
- In Texas, Jewish Republicans step in against lawsuit seeking to reject 127,000 drive-through votes - Haaretz.com - November 3rd, 2020
- Uncomfortable Conversations: Fighting racism against Asian Americans in wake of COVID - AdAge.com - November 3rd, 2020
- Suspect In Ithaca String Of Hate Crimes Arrested - WSKG.org - November 3rd, 2020
- The risk of violence and protests on Election Day - Vox.com - November 3rd, 2020
- The Color Of Water | Opinion | coronadonewsca.com - Coronado Eagle and Journal - November 3rd, 2020
- Desperation - And Response (2) - The Chattanoogan - November 3rd, 2020
- Opinion: Why Gen Z and Millennials need to vote in 2020 - WHYY - November 3rd, 2020
- ADL says election-related hate on the rise: 'antisemitism, racism, xenophobia & all forms of bigotry.' - The Boston Globe - October 25th, 2020
- CAPAC: New ADL Report Ties Spike in Anti-Asian Bigotry to Trump - The Rafu Shimpo - October 25th, 2020
- Resisting the Tide of Bigotry - The Cairo Review of Global Affairs - October 25th, 2020
- Order allowing Texas counties to have multiple mail-in ballot drop off sites is upheld, but appeal halts openings - Katy Times - October 25th, 2020
- Video of Trump rally appears to show attendee making hand gesture commonly used by white supremacists - USA TODAY - October 25th, 2020
- Hate Groups Active In Ohio Long Before Plot To Kidnap Michigan Governor - WOSU - October 25th, 2020
- Students at Foothill College pass antisemitism resolution J. - The Jewish News of Northern California - October 25th, 2020
- New Initiative Unlocks the Power of Consumers and Brings Brands and Social Media Platforms Together to Clean Up Hate Online - PRNewswire - October 25th, 2020
- Pernod Ricard gets industry buy-in for tool that reports hate speech on social media - Marketing Dive - October 25th, 2020
- 5 Things You Need to Know Today in Worcester Oct. 21, 2020 - This Week In Worcester - October 25th, 2020
- Hate Crimes on the Rise in Orange County - Fullerton Observer - October 25th, 2020
- On its 50th anniversary, the secret Jewish history of Doonesbury - Forward - October 25th, 2020
- Berkeley man known to police charged with Oakland synagogue vandalism - The Jewish News of Northern California - October 25th, 2020
- GOP mailer in the 165th decried as anti-Semitic - The Delaware County Daily Times - October 25th, 2020
- AP Explains: What's Behind Trump's Town Hall Answer on QAnon - WTTW News - October 25th, 2020
- Who are the Proud Boys: Englewood business owner proud to lead controversial group - yoursun.com - October 25th, 2020
- What Is a Militia? And Why Is the Word So Controversial These Days? - The New York Times - October 25th, 2020
- New Pernod Ricard Initiative Aims to Clean Up Hate Online - BevNET.com - October 25th, 2020
- QAnon: YouTube bans conspiracy theory content that justifies real-world violence - Euronews - October 16th, 2020
- LEADING NATIONAL AND STATE ADVOCACY GROUPS, BEN & JERRY'S, AND THE SACRAMENTO BEE ENDORSE PROPOSITION 17 - Civilrights.org - October 16th, 2020
- What You Need to Know About The Trial of the Chicago 7 - The New York Times - October 16th, 2020
- Gateway to Radicalization: Wisconsin cultivates extremism with ties to Michigan insurrection group - Milwaukee Independent - October 16th, 2020
- Affirmative Action on the Ballot: A Closer Look at Prop 16 - Fullerton Observer - October 16th, 2020
- Community organizations to hold virtual summit focused on race, hatred and social justice - SILive.com - October 12th, 2020
- Grand jury releases report on hate speech in San Mateo public schools - The Jewish News of Northern California - October 12th, 2020
- Anti-government groups shift focus from Washington to states - Idaho Business Review - October 12th, 2020
Comments