Meta’s tools increase chance of removing valuable posts about Israel-Hamas war, says watchdog

Meta’s tools increase chance of removing valuable posts about Israel-Hamas war, says watchdog
Meta’s Oversight Board expressed concerns about the removal of content that might contain evidence of human rights violations. (AFP/File)
Short Url
Updated 19 December 2023
Follow

Meta’s tools increase chance of removing valuable posts about Israel-Hamas war, says watchdog

Meta’s tools increase chance of removing valuable posts about Israel-Hamas war, says watchdog
  • The Oversight Board announced its findings after reviewing two cases on Facebook and Instagram and urged Meta to respond more quickly to changing circumstances
  • One case involved footage of the aftermath of a strike on or near Al-Shifa Hospital in Gaza City, the other the taking of hostages by Hamas on Oct 7.

DUBAI: Meta’s Oversight Board, which makes decisions about content published on the company’s platforms, published its findings on Tuesday after expedited reviews of two separate appeals from users about the removal of content relating to the war between Israel and Hamas.

The board, which completed its reviews in 12 days, expressed concerns about the removal of content that might contain evidence of human rights violations, and urged Meta to demonstrate that action was being taken to preserve such content and to respond more quickly to changing circumstances.

One appeal involved an Instagram post that showed what appeared to be the aftermath of a strike on or near Al-Shifa Hospital in Gaza City during Israel’s ground offensive. The footage showed Palestinians, including children, who had been injured or killed.

During the appeal, the creator of the post said they did not incite violence and had simply shared content that showed the suffering of Palestinians, including children, and that the removal of the post displayed bias against this.

The other case involved videos posted on Facebook of an Israeli woman begging her kidnappers not to kill her as she was taken hostage during the Oct. 7 attacks by Hamas on Israel. The creator of the post told the appeal that the video captured real events and aimed to help “stop terror” by revealing the brutality of the incidents during which hostages were taken.

The Oversight Board overturned Meta’s decisions to remove the content in both cases.

“These decisions were very difficult to make and required long and complex discussions within the Oversight Board,” said Michael McConnell, its co-chair.

Social media platforms play a critical role during times of conflict, he added, as they are often the “only vehicles” through which to “provide information, especially when the access of journalists is limited or even banned.”

Meta told the board that during the conflict in Gaza it has temporarily lowered the thresholds used by automated tools to detect and remove content that potentially violates its rules, which reduces the risk of harmful content appearing but increases the likelihood that legitimate, valuable content might be removed from its platforms. As of Dec. 11, Meta had not restored the thresholds to pre-Oct. 7 levels, the board said.

It was also revealed that there had been a near-three-fold increase in the average daily number of appeals by users relating to the Middle East and North Africa region in the weeks following the Oct. 7 attacks.

The Oversight Board highlighted four aspects of Meta’s performance it said affected freedom of expression.

When the company applied warning messages to posts to prevent the involuntary exposure of users to disturbing content, it also excluded those posts from being recommended to other Facebook or Instagram users, even in cases where it had determined that the intention of the posts was to raise awareness.

In the case of the post about the situation at Al-Shifa Hospital, the steps taken to remove the content and to reject an appeal from the user happened automatically, without any human intervention or review, resulting in the suppression of information about the suffering in Gaza, the board said.

In the case of the footage of the Israeli hostages, Meta said it initially removed the videos out of concern that they might be perceived as celebrating or promoting the actions of Hamas. A few days later, victims’ families started sharing the videos to condemn the attacks and raise awareness of the situation. The Israeli government and media organizations in the country similarly shared the footage.

Meta said it began to allow the sharing of content related to the taking of hostages on or around Oct. 20, but only by accounts subject to its Early Response Secondary Review or cross-check policy, which allows for additional reviews of content from specified accounts.

The relaxing of the rules on videos showing hostages was not expanded to include all users until Nov. 16, and even then only for content posted after that date.

The Oversight Board said that although Meta had explained the need to proceed with caution because of the “humanitarian risks of portrayals of the hostages, the company’s use of this policy highlighted concerns previously raised about unequal treatment of users.”

McConnell said: “The board focused on protecting the right to the freedom of expression of people on all sides about these horrific events, while ensuring that none of the testimonies incited violence or hatred.

“These testimonies are important not just for the speakers but for users around the world who are seeking timely and diverse information about groundbreaking events, some of which could be important evidence of potential grave violations of international human rights and humanitarian law.”

The Oversight Board also reiterated the need for Meta to “swiftly act on previously issued content-moderation guidance.”

The Oversight Board revealed on Dec. 7 that it was considering the two cases and would conduct an expedited review. This gave it 30 days to publish its findings and it completed its review in just 12 days.


OpenAI considers taking on Google with browser, the Information reports

OpenAI considers taking on Google with browser, the Information reports
Updated 22 November 2024
Follow

OpenAI considers taking on Google with browser, the Information reports

OpenAI considers taking on Google with browser, the Information reports
  • OpenAI has already entered the search market with SearchGPT
  • Google commands the lion’s share of the browser and search market

ChatGPT-owner OpenAI has recently considered developing a web browser that would combine with its chatbot and has separately discussed or struck deals to power search features, the Information reported on Thursday.
OpenAI has spoken about the search product with website and app developers such as Conde Nast, Redfin, Eventbrite and Priceline, the report said, citing people who have seen prototypes or designs of the products.
Google and OpenAI did not immediately respond to Reuters’ requests for comment.
The move could pit the Sam Altman-led company against search giant Google, which commands the lion’s share of the browser and search market. OpenAI has already entered the search market with SearchGPT.
Alphabet shares were down 1 percent after the bell, after falling nearly 5 percent in regular trading on Thursday.
 


Snap launches new office, first hub for creators in Saudi Arabia

Snap launches new office, first hub for creators in Saudi Arabia
Updated 22 November 2024
Follow

Snap launches new office, first hub for creators in Saudi Arabia

Snap launches new office, first hub for creators in Saudi Arabia
  • Diriyah’s JAX District is location for platform
  • Move will support partnership with Kingdom’s Ministry of Culture

DUBAI: Snap is expanding its presence in Saudi Arabia with the launch of a new office and the Kingdom’s first creator hub, named Majlis Snap for Content Creators, in Diriyah’s JAX District, near Riyadh.

Majlis Snap for Content Creators will serve as a platform to grow and support the local creator community through training, educational programs and opportunities for collaboration.

The opening ceremony of the new office was attended by Evan Spiegel, co-founder and CEO of Snap Inc.; Saudi Minister of Investment Khalid Al-Falih; and Saudi Minister of Communications and Information Technology Abdullah Al-Swaha.

The event featured a conversation between Spiegel and Jomana Al-Rashid, the CEO of the Saudi Research and Media Group, about Snap’s growth and popularity in the Kingdom.

Spiegel also hosted exclusive sessions with content creators to commemorate the launch of Majlis Snap for Content Creators.

The establishment of the new office will support Snap’s partnership with Saudi Arabia’s Ministry of Culture by upskilling local talent.

Abdulla Alhammadi, managing director of Snap Inc. in Saudi Arabia, said that the Kingdom was one of the company’s “most dynamic growth markets” and the investment would “enhance the experience” of both Snapchat users and advertisers.

Hussein Freijeh, the vice president of Snap Inc. in the Middle East and North Africa region, said that the company’s expansion in Saudi Arabia “symbolizes more than just a physical presence,” and represented “a deeper commitment to enhancing Saudi Arabia’s digital ecosystem.”

He added: “This marks a huge milestone in our journey in KSA, and we look forward to strengthening our connection with Saudi partners and clients, in line with the country’s digital transformation agenda.”

Snapchat has 25 million active monthly users in the Kingdom, reaching 90 percent of those aged between 13 and 34, with users opening it 50 times a day on average.


An AI-powered tool aims to combat fake news in the Arab world and beyond

An AI-powered tool aims to combat fake news in the Arab world and beyond
Updated 22 November 2024
Follow

An AI-powered tool aims to combat fake news in the Arab world and beyond

An AI-powered tool aims to combat fake news in the Arab world and beyond
  • Developed jointly with EU academic institutions, FRAPPE is the brainchild of Preslav Nakov of Abu Dhabi’s MBZUAI
  • System trained with 23 different linguistic techniques, can identify specific persuasion and propaganda techniques

RIYADH: Rising concern over disinformation’s role in manipulating public opinion has motivated Preslav Nakov, a professor at the UAE’s Mohamed bin Zayed University of Artificial Intelligence, to develop an AI-powered tool for detecting propaganda. 

FRAPPE, short for Framing, Persuasion and Propaganda Explorer, is designed to assess news framing techniques and identify potential instances of information manipulation.

Nakov, chair of the natural language processing department and professor of natural language processing at the Abu Dhabi-based MBZUAI, said that AI plays a central role in FRAPPE by analyzing, categorizing and detecting complex patterns that influence readers’ opinions and emotions. 

The tool offers real-time, on-the-fly analysis of individual articles while enabling a comprehensive comparison of framing and persuasion strategies across a wide range of media outlets, he told Arab News. 

The UN defines disinformation as inaccurate information deliberately created and disseminated with the intent to deceive the public and cause serious harm. It can be spread by both state and non-state actors and can affect human rights, fuel armed conflict and undermine public policy responses. 

The Global Risks Report 2024 by the World Economic Forum identifies misinformation and disinformation as a top short-term global risk. These forms of deceptive communication not only mislead the public but also erode trust, deepen societal divisions and threaten fundamental human rights. 

Nevertheless, the WEF highlighted in an article in June that while AI technologies are being used in the production of both misinformation and disinformation, they can be harnessed to combat this risk by analyzing patterns, language and context. 

Prof. Preslav Nakov, developer of an AI-powered tool for detecting propaganda. (Supplied)

Nakov said that FRAPPE, trained with 23 different linguistic techniques, “uses AI to identify specific persuasion and propaganda techniques, such as name-calling, loaded language, appeals to fear, exaggeration and repetition.” 

“FRAPPE further uses AI to perform framing analysis,” he said, adding that the tool distinguishes “the main perspectives from which an issue is being discussed: Morality, fairness, equality, political, and cultural identity.” 

With a database of in excess of 2.5 million articles from more than 8,000 sources, the multilingual system enables users to explore and compare how different countries and outlets frame and present information. 

DID YOUKNOW?

• Disinformation is the intentional spread of false information to sway public opinion.

• Propaganda often employs loaded language to elicit emotional reactions.

• A WEF report identifies disinformation and misinformation as a top short-term risk.

Moreover, to build the training data for the system, more than 40 journalists from several European countries contributed to the manual analysis of news content in 13 languages. 

This manual analysis, according to Nakov, allows FRAPPE to discern the underlying frames that shape how stories are told and perceived. By identifying the dominant frames within an article, FRAPPE compares these across media sources, countries and languages, providing valuable insights into how framing varies globally. 

FRAPPE is designed for a broad audience, including the general public, journalists, researchers, and even policymakers. 

With an extensive database, FRAPPE's  multilingual system enables users to explore and compare how different countries and outlets frame and present information. (Supplied)

“For the general user, FRAPPE serves as an educational tool to explore how news content is framed, enabling them to identify propaganda techniques like name-calling, flag-waving, loaded language and appeals to fear,” Nakov said. 

“For journalists and policymakers, FRAPPE offers a powerful tool to examine and compare framing and persuasion strategies across different countries, languages and outlets,” he added. 

The system relies on annotations from journalists who manually identified persuasion and propaganda techniques across a wide range of articles. This minimizes the risk of overly subjective or one-sided interpretations. 

Opinion

This section contains relevant reference points, placed in (Opinion field)

Transparency and unbiased analysis were fundamental in the development of FRAPPE. Nakov said: “Users should be aware that our models use neural networks and, as such, they lack explainability.” 

He also warned that “despite our intent, due to potential unintended article selection biases, FRAPPE might be favoring some political or social standpoints.” 

On the positive side, however, “FRAPPE has the potential to influence the way news articles are perceived and consumed, and journalists may become more aware of the language they use and its potential impact on readers.” 

FRAPPE has the ability to spot persuasive or manipulative techniques in news content. (Supplied)

To spot persuasive or manipulative techniques in news content, Nakov advises readers and viewers to “watch out for emotional language designed to provoke strong reactions like fear or anger, and be mindful of loaded words, such as ‘radical’ and ‘heroic, which carry emotional weight.”

He urged readers to critically assess articles that rely too heavily on a single expert or selective quotes, stressing the importance of considering how different outlets might report the same event in contrasting ways. 

To gain a clearer perspective, Nakov advises cross-checking sources and comparing how different media outlets cover the same story. This approach helps reveal varying angles, biases or framing techniques. 

He also stressed that oversimplified “us versus them” narratives “often indicate manipulation, as do articles that frame an issue with a particular angle, leaving out important details. 

FRAPPE has been featured in numerous EU workshops focused on combating fake news. (Supplied)

“False dilemmas, where only extreme choices are presented and repetitive phrases meant to reinforce a point are also red flags,” he said. 

“FRAPPE envisions empowering individuals and institutions to make more informed decisions by revealing the framing and persuasion techniques embedded in media content. Its aim is to enhance transparency in journalism, promote trust in media and contribute to a more informed, media-literate public.” 

Developed in collaboration with the European Commission’s Joint Research Center and several academic institutions across Europe, FRAPPE was launched ahead of the 2024 European Parliament election, held in the EU between June 6-9 this year. 

The tool, integrated into the Europe Media Monitor, has been featured in numerous EU workshops focused on combating fake news.


 

 


Israel faces backlash for defacing Lebanese war memorials

Israel faces backlash for defacing Lebanese war memorials
Updated 21 November 2024
Follow

Israel faces backlash for defacing Lebanese war memorials

Israel faces backlash for defacing Lebanese war memorials
  • At least 2 memorials honoring victims of Hula massacre targeted
  • Incidents are latest in series of alleged attacks on cultural, religious heritage sites

LONDON: Israel’s reported defacing of war memorials in Lebanon has ignited widespread outrage online, with critics accusing Tel Aviv of yet another “immoral” act during its ongoing conflict.

“A memorial in the village of Hula, commemorating the massacre committed by the Israeli army in 1948, defaced & desecrated with the above Hebrew graffiti, by Israel’s most immoral army in the world,” said Lebanese-British journalist and author Hala Jaber on X.

According to online reports, which Arab News could not independently verify, Israel’s Golani Brigade allegedly vandalized a memorial in Hula — a southern Lebanese village — dedicated to victims of a 1948 massacre. Graffiti sprayed on the memorial reportedly read: “A good Shiite is a dead Shiite.”

The desecration has drawn sharp criticism, with users on social media highlighting the act as emblematic of broader issues within Israeli society.

A user said: “Netanyahu represents a large part of Israelis … no, the war and the atrocities committed by Israel are not only the work of Netanyahu … the evil of Israeli society is much deeper.”

L’Orient-Le Jour quoted Hula City Council Chairman Chakib Koteich confirming the vandalism, as well as the destruction of a monument commemorating the same massacre.

Funded by Lebanon’s Southern Council, the memorial was unveiled in 2002 in a ceremony led by Parliament Speaker Nabih Berri.

The memorial honored victims of the Oct. 31, 1948, Hula massacre, in which members of the Haganah, a Jewish paramilitary group, disguised as Arab Relief Army soldiers, surrounded and attacked the village.

Over two days, 80 residents — men, women, children, and the elderly — were killed and the village’s 250 homes destroyed.

The incidents are the latest in a series of alleged attacks on cultural and religious heritage sites in Lebanon and Gaza since October 2023.

Local media reported in November that Israeli forces had destroyed cemeteries and historic burial sites in southern Lebanon, including the ancient shrine of Prophet Benjamin in the village of Mhaibib.

Israeli forces were earlier accused of demolishing a memorial to Palestinian journalist Shireen Abu Akleh at the entrance to the Jenin refugee camp in the occupied West Bank.

UNESCO recently placed 34 Lebanese heritage sites under “enhanced protection,” citing targeted strikes near World Heritage landmarks in Baalbek and Tyre. These areas, both Iran-backed Hezbollah strongholds, are home to ancient Roman ruins of global cultural significance.

The outrage comes as the International Criminal Court on Thursday issued arrest warrants for Israeli Prime Minister Benjamin Netanyahu and other officials, accusing them of war crimes in Gaza.

While the suspects are unlikely to face trial, the announcement could alter the dynamics of the current conflict, although its broader implications remain uncertain.


Tuwaiq Academy becomes authorized Google Cloud training partner

Tuwaiq Academy becomes authorized Google Cloud training partner
Updated 21 November 2024
Follow

Tuwaiq Academy becomes authorized Google Cloud training partner

Tuwaiq Academy becomes authorized Google Cloud training partner

RIYADH: The leading Saudi programming trainer Tuwaiq Academy has been selected as a Google Cloud Authorized Training Partner to introduce boot camps and programs featuring professional certifications.

A number of the academy’s staff members have received certification in instructing Google Cloud’s cloud computing technologies and services.

Google Cloud’s endorsement of Tuwaiq Academy highlights the institution’s adherence to worldwide standards in fostering expertise in cloud infrastructure, data science, machine learning and application development while providing professional certifications, said CEO of Tuwaiq Academy Abdulaziz Alhammadi.

These certifications include certified professional cloud architect, certified professional data engineer, certified professional cloud developer, certified professional cloud security engineer and certified professional machine learning engineer.

This milestone follows the staff’s acquisition of various professional certifications in teaching cutting-edge technologies across multiple cloud computing disciplines.

Alhammadi highlighted the academy’s dedication to forging partnerships with prominent global organizations to offer professional boot camps and programs within an environment equipped with the latest technologies.

The objective is to cultivate outstanding national talents capable of developing innovative solutions across diverse sectors.

Tuwaiq Academy stands out as the first of its kind to offer a multitude of boot camps and programs in partnership with leading global entities, benefiting more than 1,000 trainees daily, Alhammadi said.

Founded in 2019, it provides a range of training and educational courses in cybersecurity, programming and software development in a bid to position Saudi Arabia among the ranks of technologically advanced countries.

The academy employs a practical application-based learning methodology to remain current with modern technological advancements and align with job market demands.