COVID-19: Dealing with Gaps in the Data

One of the major problems with the COVID-19 pandemic is the speed at which the contagion spreads. This makes treating infected people much more difficult to manage, but also severely hinders our ability to have an up-to-date, thorough and trustworthy picture of the situation in Europe and the rest of the world.

The information we rely on is approximate and often errs on the side of caution (for example, the number of infected people, or deaths caused by the pandemic). It’s important to be aware of these limitations, and approach the data with caution, even if this data is the best we have, given the present circumstances. Of all official data on the global situation, that produced by the European Centre for Disease Prevention and Control (ECDC) is considered among the most reliable. Nevertheless, new and more accurate studies are emerging every day, providing additional data to help understand the pandemic and its course of development.

How many are really infected?

We don’t know. What we do know is the number of confirmed infections – individuals testing positive for the virus – and highly approximate estimates of total infections.

The test for the virus involves taking a sample of saliva or mucus, which is then analysed for traces of the virus’ genetic code. The number of people being tested varies widely from country to country: depending, above all, on how well-equipped a country is to perform large-scale testing (often it’s not the kits that are lacking, but the personnel and laboratories required to analyse huge quantities of swabs). In certain countries, authorities decide to focus on people already showing symptoms associated with COVID-19, or even just those who are already hospitalised. We know, however, that many who have contracted the virus do not show any symptoms, or only start to show symptoms many days after being infected.

The percentage of infected people accounted for in the data varies widely from country to country. This makes it difficult to compare the development of the pandemic in different times and places. For example, Italy has performed around 3500 tests for every million inhabitants, compared to 6100 in South Korea, and 600 in Spain. According to an estimate attempted by the Centre for the Mathematical Modelling of Infectious Diseases, in the London School of Hygiene & Tropical Medicine, Italy and Spain may have only recorded 5 percent of people actually infected. 

How many have really died?

This is also unknown, even if the number of deaths can be estimated with more precision than cases of infection.

What we do know is the number of deaths attributed to COVID-19 (unfortunately, the criteria for attribution are not yet internationally standardised). However, we cannot be sure that all deaths caused by the coronavirus have been recorded: in the most heavily hit areas of Italy, indications suggest that tests are not performed on all victims (many of those who die at home or in retirement homes, for example). Moreover, authoritarian regimes such as China and Iran may have an interest in publishing incomplete data in order to downplay the severity of the problem – thus the number of deaths caused by the pandemic may very well be higher than suggested by official counts.

How deadly is COVID-19?

No certainty here either. The relative danger of a disease can be measured by its case fatality rate – the number of deaths as a proportion of those infected – or the mortality rate, which measures the number of deaths as a proportion of the population. A case fatality rate of 4 percent indicates that for every 100 people infected the disease causes an average of four deaths.

The available estimates of COVID-19’s case fatality rate vary all too widely according to context. On the one hand, such variations could in fact be tied to local factors: for example, the disease is likely to have a greater impact in regions or countries where the population is older or more prone to respiratory illnesses, such as heavily polluted Northern Italy. Alternatively, such variations may only be apparent, and caused by differences in how data is collected. The case fatality rate compares two figures – deaths and infections – but, as we have seen, these figures are often recorded in different ways, and often contain significant gaps.

In any case, COVID-19’s case fatality rate is an order of magnitude greater than that of more mundane viral illnesses, such as seasonal flu. The latter typically causes the death of fewer than 0.1 percent of people infected, over many months, while it is estimated that COVID-19 causes an at least twenty or thirty times higher percentage of deaths, over just a few weeks. 

Two useful techniques for comparing data

Apart from the gaps and disparities in data collection, comparisons between regions and countries affected by the coronavirus are complicated by the fact that contagion didn’t start everywhere at the same time. Comparing Hubei province in China – where infection began around a month ago – with a country where contagion has just begun would not be particularly instructive. In order to compare such contexts, we should start with the day when the outbreak was registered in each area, and compare developments from there. For example, 15 days after the virus broke out in Italy, around 800 deaths had been recorded there, while in Spain, 15 days after the virus was detected in its territory, 2000 deaths had been recorded.

Another way to compare developments in countries with different data collection methods is to compare the rates of contagion in each country – for example, measuring the number of days it took for the number of confirmed deaths to double. In Germany, the figure doubled every two days, and in Italy every five days. In South Korea it has taken 13 days for the number of confirmed deaths to double, indicating that contagion has slowed down considerably.

Facebook Extends Political Ads Rules to Balkans Before Elections

Facebook has added Montenegro, North Macedonia, Serbia, Moldova and Turkey to its updated list of countries that must adhere to its strict political advertising transparency regulations.

The enforcement of the new regulations is expected in mid-March, about a month before parliamentary elections in Serbia and North Macedonia, due on April 26 and April 12 respectively. Montenegro is also due to hold the elections this year, by October at the latest.  

Facebook launched the regulations in June last year. They mean that any adverts paid for by a political group or candidate must be labelled as such. 

With the new 32 countries, the total count of states required to stick to the regulations, which relate to any adverts about social issues, elections or politics, rose to 89.

Apart from the five Balkan countries, the regulations will also expand to Chile, Japan, Mexico and Indonesia, to name a few. Facebook said it was working to expand enforcement to more countries later this year, including Myanmar and Brazil.

“Anyone who wants to run ads about elections or politics in these countries will need to confirm their identity with an ID issued from the country they want to run ads in and disclose who is responsible for the ad.

“We require that the advertiser provide additional information, like a local business address, local phone number, email and website, if they choose to use their organization or Page name in the disclaimer. These requirements hold advertisers accountable for the ads they run on Facebook and Instagram,” the social media giant wrote on its Facebook for Business website

While political ads will be more transparent, the requirements are also designed to ensure that Facebook can list political advertisers in its Ad Library. Political ads from all of these new regions will now be added to the Ad Library API.

“We will introduce the Ad Library Report for each of these countries by the end of April. The report provides aggregated insights for ads about elections and politics, such as total number of ads and spends in the Ad Library. The report is also available as a downloadable file,” Facebook explained.

Political parties in the Balkans have often used social media in ways that are far from transparent during elections.

At the beginning of 2020, SHARE Foundation, a Serbia-based digital rights NGO, asked Facebook to put Serbia and North Macedonia on the list, citing the upcoming elections as the main reason. In its letter, SHARE said that if these two countries were on the list “the campaign will be more transparent.”

“This is especially important, since it is expected that Facebook pages which are not openly political might engage in supporting a certain political party or candidate with ads,” the letter adds.

French Court Rules against Facial Recognition in High Schools

A court in Marseille ruled on Thursday that authorities in France’s southeastern Provence-Alpes-Cote d’Azur region had no power to authorise the use of facial recognition systems in two high schools in Nice and Marseille.

The city’s Administrative Court overturned the decision of regional authorities, ruling that only schools had the power to authorise such technology.

The court ruled that the decision breached the General Data Protection Regulation, GDPR, as such systems are based on consent but students cannot give consent freely given the relationship of authority that binds them to the school’s administration.

“To my knowledge, this is the first judgment in France concerning the use of facial recognition technologies in public space,” said Alexis Fitzjohn O Cobhthaigh, a lawyer representing several associations that brought the case to court.

Disproportionate measure

The case stems from an experiment launched at the end of 2018 to equip the Ampère high school in Marseille and Les Eucalyptus in Nice with virtual access control devices, by which cameras would recognise high school students and grant them access and be able to follow the trajectory of people.

A number of digital and human rights organisations said the plan violated individual freedoms. France’s National Data Protection Commission, CNIL, also came out against it in October 2019, calling the experiment disproportionate and illegal.

“This installation cannot be implemented legally,” the head of the CNIL wrote to the regional authority in charge of approving the trials, according to a letter cited by the investigative website Mediapart.

According to French media, parents and teachers’ unions also opposed the experiment.

The Administrative Court ruled that using facial recognition to control access to high schools was a disproportionate measure.

Nevertheless, some French media reports said regional authorities were pressing ahead with the plan regardless of the court’s ruling.

Call for total ban

The case was brought in February 2019 by French advocacy group La Quadrature du Net, which works to promote and defend fundamental freedoms in the digital world.

“In France, this is the first court decision about facial recognition and the first success against it! We hope it will be followed by a series of other successes leading to the total ban of facial recognition,” the group wrote on their website on February 27.

La Quadrature du Net and 80 other civil society groups signed a joint letter on December 19 calling on French authorities to ban facial recognition for any purposes of security and surveillance, citing similar bans in San Francisco and other US cities.

“Facial recognition is a uniquely invasive and dehumanising technology, which makes possible, sooner or later, constant surveillance of the public space,” they wrote.

“It creates a society in which we are all suspects. It turns our face into a tracking device, rather than a signifier of personality, eventually reducing it to a technical object. It enables invisible control. It establishes a permanent and inescapable identification regime.”

EU Drafts ‘Human-Centric’ AI Plan to Match US, China

The European Commission on Wednesday unveiled the white paper as a part of a European digital strategy on developing artificial intelligence, designed to compete with US and Chinese sector leaders while also addressing potential human rights abuses associated with this emerging technology.

“Europe’s digital transition must protect and empower citizens, businesses and society as a whole,” European Commission President Ursula von der Leyen wrote in an op-ed that outlined the key points of the proposed blueprint.

“To make this happen, Europe needs to have its own digital capacities – be it quantum computing, 5G, cybersecurity or artificial intelligence,” Von der Leyen explained.

She said the Commission should make available the necessary funding to “draw in national and private sector funds” to develop these technologies within the EU, and ensure what she called “tech sovereignty” for the bloc.

According to the white paper, investment in artificial intelligence will be channelled through the Horizon Europe programme, which is to be allocated 15 billion euros in the coming 2021-2027 Commission budget.

The white paper provides also for further investment in adopting new legislation and building safe data spaces, in order to consolidate the EU’s leading role in data protection and assure “the development of AI in Europe whilst ensuring respect of fundamental rights”.

The cornerstone of the new legislation, to be gradually enforced in the EU space, the draft says, might be the Ethics Guidelines for Trustworthy AI. This is a set of recommendations drawn up by a panel of experts that was tested by companies in 2019.

The proposed strategy aspires to promote “a human-centric approach” to AI in line with “European values”. In order to ensure that, the paper advocates tough legislation to counter the risks to human rights of some of the more “intrusive” applications of AI, such as facial recognition and its use for remote identification.

Facial recognition is currently banned in the EU. The white paper aims to promote a “broad debate on which circumstances might justify exceptions in the future, if any,” the Commission noted in a statement.

Moreover, the document commits to putting in place a mechanism capable of identifying and banning any AI algorithms used in “predicting criminal recidivism” that “can display gender and racial bias, demonstrating different recidivism prediction probability for women vs men or for nationals vs foreigners”.

The white paper pledges to ensure that victims of abuse of artificial intelligence and other digital technologies do not encounter any more difficulties in getting compensation than victims of abuses of more traditional products and services.

The document also presents a proposed European Data Strategy, harmonized with the existing General Data Protection Regulation and intended to “create a genuine single market for data, where personal and non-personal data … are secure and where businesses and the public sector have easy access to huge amount of high quality data to create and innovate”.

Selling .ORG Puts Civil Society at Risk

Executive directors of 11 international NGOs released an open letter calling on the leaders of Internet Society (ISOC) and Internet Corporation for Assigned Names and Numbers (ICANN) to stop the sale of the .ORG top-level domain to private equity firm Ethos Capital. 

“.ORG is the place where civil society and NGOs reside in the digital environment.  Both the physical and virtual world have become increasingly inhospitable and risky for civil society organizations who face constant surveillance, online censorship, and even more physical risks and legal restrictions on their operations and personnel. This proposed sale presents an additional danger to civil society and undermines the safety and stability of the digital space for countless non-governmental organizations, their partners, and their broader communities,” the letter reads.

Signers include the directors of Greenpeace International, Human Rights Watch, International Trade Union Confederation, Amnesty International, 350.org, Transparency International, Access Now, Sierra Club, American Civil Liberties Union (ACLU), Consumer Reports, and Color of ChangeThe letter is being officially released in Davos at the World Economic Forum, where global business, government, and social leaders are gathered to discuss priorities for 2020 and beyond. 

“Free expression around the world is increasingly endangered by government and corporate players, which is why we are joining other civil society organizations in making public our concerns over the .ORG sale,” said Anthony D. Romero, Executive Director of the American Civil Liberties Union (ACLU). “The internet is crucial to the integrity of civil liberties and human rights work, and also the safety of those doing it. The security of civil society should not be entrusted to private equity.”

“Even more so than what .ORG would look like in the next five years, I’m deeply worried about its fate in 2040,” said Brett Solomon, Executive Director of Access Now. “If .ORG is transferred to the private sector, it would inevitably make its way into the hands of those who stand to gain from its control and are willing the pay the price to have it — that could be, for example, the Saudi or Chinese government, or surveillance tech investors like Novalpina Capital.”

There has been a resounding rejection of the sale from the .ORG community and other concerned stakeholders around the world, in particular due to the lack of transparency around the deal and the absence of safeguards for the domain’s continued stability, security, and accessibility.  Nearly 700 organizations and over 20,000 individuals have signed on to the SaveDotOrg petition calling to stop the sale.

EDRi Publishes Guide for Ethical Website Development and Maintenance

European Digital Rights, EDRi, released the new guide for ethical website development and maintenance, Ethical Web Dev.

The guide is aimed at web developers and maintainers who have a strong understanding of technical concepts, to assist them in bringing the web back to its roots – a decentralised tool that can enhance fundamental rights, democracy and freedom of expression.

The goal of the project, which started more than a year ago, was to provide guidance to developers on how to move away from third-party infected, data-leaking, unethical and unsafe practices.

The guide is a result of an extensive collective work, with inputs from experts of the EDRi network (Anders Jensen-Urstad, Walter van Holst, Maddalena Falzoni, Hanno “Rince” Wagner, Piksel), external contributions (Gordon Lennox, Achim Klabunde, Laura Kalbag, Aral Balkan), and the crucial help of Sid Rao, Public Interest Technologist and ex-Ford-Mozilla Fellow at EDRi.

The guide is distributed under a Creative Commons 4.0 Licence.

Download:

Ethical Web Dev – Guide for ethical website development and maintenance
https://edri.org/files/ethical_web_dev_web.pdf

Alexander the Bot: The Twitter War for the Macedonian Soul

As the sun dipped over the rooftops of the North Macedonia town of Bitola near the southern border with Greece, the man they call “Cheese” sipped a beer on the Sirok Sokak pedestrian strip.

As sundowns go, this one seemed fitting. It was August 12, the day North Macedonia outlawed the use of the Vergina Sun – a Greek national symbol – in books, on monuments and in public spaces.

For Cheese, the ban on “appropriation” of the Classical Hellenic emblem with its distinctive pointy rays was the latest act of surrender in a bitter fight over Macedonian identity.

It was part of a historic deal with Greece to end a 30-year dispute over his country’s use of the name “Macedonia” – which Athens argued implied territorial ambitions over a northern Greek province of the same name and its ancient legacy of Alexander the Great.

Under the deal signed in July 2018, the former Yugoslav republic had to change maps and textbooks, abandon all use of the Vergina Sun and – the ultimate betrayal, in Cheese’s view – rechristen itself “North Macedonia”.

Sitting in an outdoor cafe as dusk descended, he vowed never to sully his lips with the new name.

“I’m a patriot, and I just don’t want my country’s name to be changed,” he told BIRN.

Few people know Cheese’s true identity, though many are familiar with his nationalist views. He is, in fact, Goran Kostovski, a 38-year-old marketing company worker from the capital, Skopje.

With almost 10,000 Twitter followers on three continents, Kostovski led a social media campaign in 2018 urging Macedonians to boycott a referendum on implementing the name-change deal, known as the Prespa agreement after the lake near which it was signed.

While the Prespa deal promised to unblock Greek opposition to the country’s hopes of joining NATO and the EU, critics saw it as a compromise too far. They hoped a low turnout in the September 2018 referendum would invalidate the result.

“It made no sense to tell the world to vote no in the referendum because we feared the government would distort the results,” Kostovski said. “We had to boycott the referendum first.”

Prompting street protests at home and drumming up diaspora dollars abroad, the “#boycott” campaign was a runaway success.

While 95 per cent of those who voted in the referendum were in favour of the name-change deal, turnout was only 37 per cent – well short of the 50 per cent minimum threshold.

Though parliament later ratified the Prespa agreement anyway, experts say the victory for voter suppression was due in part to a new type of information warfare increasingly seen in nationalist circles.

Known as “computation propaganda”, it is what the Oxford Internet Institute at Oxford University calls “the interaction of algorithms, automation and politics”.

Few have mastered the art better than Kostovski, though he is cagey about the methods he uses.

“You can say we’re bots, but that doesn’t mean it’s true,” he said, referring to the new foot soldiers of the online propaganda war: bogus Twitter accounts programmed to behave like humans.

“We’ve blurred your thinking so you don’t know where our campaign is coming from, and you don’t know where to look first.”

While much has been said of Balkan troll farms and fake news factories, less is known about the impact of computational propaganda on the workings of democracy in the region.

A BIRN investigation into nationalist networks on both sides of the name dispute lifts the lid on the online tricks employed to amplify political messages and distort public opinion.

It is a journey into an underworld of computer code and conspiracy theories, where “ghost users” and “Twitterbots” meet far-right extremism in a digital hall of mirrors.

As much fake buzz as fake news, the activity is designed to create the false impression of a giant online conversation so opinion-makers such as journalists and activists sit up and take notice.

In this way, experts say a small group of geeks with laptops can exert an influence way out of whack with their actual numbers, with worrying implications for democratic discourse.

Goran Kostovski, the man behind the ‘Cheese’ account on Twitter, sits in an outdoor cafe in the North Macedonia town of Bitola. Photo: Kostas Zafeiropoulos

Disinformation ‘spin cycle’

At the government headquarters in Skopje, the country’s new official name – Republic of North Macedonia – greets visitors as they approach the Ionic columns of the building, renovated five years ago to look like the White House in Washington, DC.

It is a stone’s throw from the city’s main square, where a statue of Alexander the Great on a stallion looms over a Classical-style fountain – the result of a taxpayer-funded makeover of Skopje to give it a more antiquarian feel.

Many saw the revamp announced in 2010 as an architectural thumbing of the nose at Greece by the government of then Prime Minister Nikola Gruevski after Athens vetoed his country’s accession to NATO in 2008.

Inside government headquarters, Demijan Hadzi-Angelovski, a 28-year-old social media expert at the information ministry, recalled how 10 or so influential Twitter accounts sought to dominate the news agenda in the run-up to the Prespa referendum.

Every day, three times a day, a different user would send one or two provocative tweets, which would then be liked and retweeted by an army of automated accounts, he said.

The idea was to “trend” on Twitter and get picked up by big news aggregators like Time.mk.

“Their goal was to have the news sites view and reproduce these tweets, to make the information more credible,” he said. “They then re-posted the news in a washing machine news cycle.”

Their goal was to have the news sites view and reproduce these tweets, to make the information more credible. They then re-posted the news in a washing machine news cycle.

Demijan Hadzi-Angelovski, government social media expert

According to Information Minister Damjan Manchevski, who oversaw the government’s pro-Prespa referendum campaign, much of the recycled content was fake news designed to discredit the agreement.

“More than 10 per cent of articles in that period were pure misinformation,” Manchevski told BIRN in an interview. “The bots on Twitter were the main source of fake news.”

Fireworks light up the sky behind a statue of Alexander the Great unveiled in Skopje’s main square in September 2011. The installation of the sculpture was part of a cultural project called ‘Skopje 2014’ designed to give the city a more Classical look. EPA/NakeBatev

One story falsely stated that people living near the country’s largest army base in the central Krivolak region would be poisoned by depleted uranium brought in for military training if the government ratified the Prespa deal and then joined NATO.

An investigation by the Organised Crime and Corruption Reporting Project and the Investigative Reporting Lab Macedonia (IRL) traced the story to Zlatko Kovac, a 50-year-old US-Macedonian who works as a Washington columnist for Russian news agency Sputnik.

Kovac did not reply to BIRN’s request for comment.

“Kovac collaborates with a number of websites in [North] Macedonia that are part of the propaganda mechanism against the Prespa Agreement,” said Saska Cvetkovska, the chain-smoking editor-in-chief of IRL.

“The story was started by Kovac on Facebook, the news was immediately posted on Twitter, dozens of bots reproduced it and then several conservative online media … posted it as a regular news item.”

As a result, Defence Minister Radmila Sekerinska spent a week frantically reassuring people it was not true, Cvetkovska said.

In the days before the referendum, other scare stories wormed their way into mainstream news.

Media reported that people could be prosecuted for disagreeing with Prespa, that the need to print new money would cause massive inflation and that Greece would get a blank cheque to do whatever it wanted.

None of this happened by accident.

Anti-Prespa demonstrators protest in front of the parliament building in Skopje in November 2018. Photo: EPA-EFE/GEORGI LICOVSKI

‘Cyborg bots’

In the runup to the referendum, the Transatlantic Commission on Election Integrity, an initiative of the Danish-based Alliance of Democracies Foundation, noticed a spike in bot activity on its social media monitoring tools.

“There is clearly a concerted effort to thwart the democratic rights of Macedonians and delegitimise the referendum vote,” it said in a statement.

There is clearly a concerted effort to thwart the democratic rights of Macedonians and delegitimise the referendum vote.

Transatlantic Commission on Election Integrity

Two weeks before the referendum, the Digital Forensic Research Lab (DFRLab) of the US-based Atlantic Council think tank, which monitors digital propaganda, published research showing that far-right Twitter accounts were boosting the boycott campaign.

Over a period of nine days, researchers analysed all tweets linked to the hashtags “#Бојкотирам” and “#bojkotiram” (#boycott) – around 23,800 of them.

They found that more than 80 per cent were in fact retweets, a  4:1 ratio of retweets to original content that suggested rampant automation.

“There was well-coordinated, non-authentic activity that destroyed any normal Twitter talk,” DFRLab researcher Kanishk Karan told BIRN. “Instead of discussing these accounts, they spammed others and bombarded them with thousands of mentions and retweets.”

The DFRLab identified the nine most active Twitter accounts that helped the campaign go viral – and Kostovski’s “Cheese” account was among them.

A DFR Lab visualisation shows the most influential accounts around the #boycott campaign. ‘Cheese’ is ‘C4i72’. Image courtesy of DFRLab

According to Kostovski, the campaign had three main ringleaders: himself, a blogger friend named Igor Pipovski (whose Twitter handle “@m0rban” honours populist Hungarian Prime Minister Viktor Orban) and Zarko Hristovski, a Macedonian web developer who lives in Norway and built the campaign’s website.

#Бојкотирам belongs to the Macedonian people,” Pipovski tweeted to BIRN when asked about his role in the campaign. “Nobody should take the credit.”

BIRN was unable to contact Hristovski but Kostovski expounded on the worldview that he said motivated them all.

“We live in the middle of a digital war,” he told BIRN. “We nationalists and patriots on the one side, and internationalists, communists and former communists and social democrats on the other.”

We live in the middle of a digital war. We nationalists and patriots on the one side, and internationalists, communists and former communists and social democrats on the other.

Goran Kostovski, alias ‘Cheese’

His Cheese account – which is followed by one in three Twitter users in North Macedonia, according to analysis using the Statcounter online monitoring  tool – makes no bones about its purpose.

A pinned tweet at the top of his Twitter stream says: “A place where you will find plenty of banners, gifs, memes and other propaganda material that will be useful for a successful social campaign. Network against the fatal referendum to change our identity.”

In many ways, Kostovski has as much in common with alt-right white supremacists in the United States as with Macedonian nationalists. His Twitter posts bristle with far-right symbols and conspiracy theories.

His Twitter profile picture was formerly a cartoon image of Pepe the Frog, a favourite alt-right emblem. These days, his header has a big “Q”, a reference to the popular far-right “QAnon” conspiracy theory of a “deep state” plot against US President Donald Trump.

In a rare public appearance, Kostovski addressed a rally in Skopje two days before the referendum wearing a “Q” hat and a “Make America Great Again” T-shirt.

“We all fight the deep state and globalisation,” Kostovski told BIRN. “We have the same enemies and similar ideologies.”

He listed the populists who inspire him: Trump, Orban, French far-right opposition leader Marine Le Pen and British Brexit Party founder Nigel Farage. Many of his Twitter posts also feature former Trump strategist Steve Bannon.

“We were excited about Trump’s election campaign and we tried to copy its methods and symbols,” Kostovski said. “We wanted people especially to believe that Steve Bannon was involved in the [#boycott] campaign.”

In one tweet in August 2018, he suggested that the “illegal and treasonous” government of Prime Minister Zoran Zaev was in hock to US billionaire philanthropist George Soros – a familiar figure of hate among far-right groups.

“Boycott the illegal #Referendum for the Nazi #agreement,” he added.

#SorosGovernment is illegal and treasonous. We will #LockThemUp for high treason. Soon… #Бојкотирам Boycott the illegal #Referendum for the Nazi #agreementpic.twitter.com/ori2aoATxe

— ︽☆︽ Чиз™ ︽☆︽ #АΨ (@C4i7Z) August 22, 2018

Kostovski insisted that he and his fellow keyboard crusaders were in no way out of the ordinary.

“In our group, we are all just normal people, with regular jobs and families, and most of us want to remain anonymous on social media.”

But BIRN’s investigation shows there is more to Kostovski’s anonymity than meets the eye.

Using a crawler application called Twitterbots, a tool created by Athens-based software engineer Dimitris Papaevangelou to assess the likelihood of bot activity, BIRN analysed Cheese’s Twitter output and found he averages almost 110 “actions” per day.

Computer scientists say any number of actions – tweets, retweets, likes and other interactions – over 70 suggests bots are on the loose.

While Kostovski said he only used automation in the early days of the #boycott campaign, BIRN’s analysis confirmed that his Cheese persona is what is known in the computational propaganda business as a “cyborg bot” – half person, half machine.

These hybrids combine algorithmic automation with human intervention to get past Twitter’s anti-bot defences, since automation is strictly no-go on the social media platform.

The Twitterbots crawler application shows Cheese’s network. Every node represents a different Twitter account. Red nodes are assumed to be bots or cyborg bots, also known as ‘retweeters’ since they combine real human interaction with automation. Image: Twitterbots screengrab

Ben Nimmo, a digital propaganda specialist at the Atlantic Council, has described the use of such bots as “a game of numbers”.

“If you create a sufficient number of false accounts and automate them, then there is a chance that they appear on the list of trending subjects,” he told this reporter in an interview for a recent investigation by the Athens-based Mediterranean Institute for Investigative Reporting (MIIR).

“Social media is therefore the place where, with the proper tactics and five persons, you can generate the impression that five million people are talking about something.”

Social media is the place where, with the proper tactics and five persons, you can generate the impression that five million people are talking about something.

Ben Nimmo, digital propaganda at the Atlantic Council

Scrutiny of Cheese’s network using Sparktoro analytical software revealed that more than 38 per cent of his followers were classified as “fake” – likely to be bots or other tools of computational propaganda.

Asked what it was doing to counter such activity, Twitter referred BIRN to its recently updated policy against what it calls “platform manipulation”, which includes spam, “malicious automation” and the use of fake accounts.

The company noted that in May 2018, it identified and challenged more than 9.9 million potentially “spammy” or automated accounts. In September, however, it reported “a nearly 50 per cent drop in challenges issued to suspected spam accounts compared to the previous reporting period”.

The website of the ‘#boycott’ campaign. The headline reads: ‘Macedonia has its own name.’ Photo: Kostas Zafeiropoulos

Diaspora dollars

According to Kostovski, nationalist politicians were quick to jump on the #boycott bandwagon in opposing the Prespa deal.

Among them was Filip Petrovski, a former lawmaker with the right-wing opposition VMRO DPMNE party who was involved in “Macedonia Boycotts”, a coalition of almost 30 small right-wing parties, political factions and civic associations.

Kostovski said the two met in the summer of 2018 to discuss working together. Contacted by BIRN, Petrovski confirmed that he was actively involved in the boycott campaign.

As the main opposition party, VMRO DPMNE’s official position towards the referendum was that people should vote with their conscience, though critics say the nationalist wing of the party was firmly against the Prespa deal.

“We have strong evidence that the centres of these online attacks are linked to the VMRO opposition, but they were not the strongest,” Information Minister Manchevski told BIRN.

Rather, he said the most strident opposition to Prespa came from Macedonians living abroad. He cited the example of a Toronto-based businessman named Bill Nikolov, president of Macedonian Human Rights Movement International in Canada.

“The most extreme of the diaspora, like him, are second-generation immigrants who have come to the country only a few times for vacations,” Manchevski said.

Kostovski said Nikolov funded an anti-Prespa billboard campaign in Skopje after getting fired up online.

“Many rich diaspora people with connections saw what we were doing on social media and multiplied our influence,” he said. “Bill Nikolov was one of them.”

Many rich diaspora people with connections saw what we were doing on social media and multiplied our influence.

Kostovski

Asked to comment, Nikolov told BIRN in a Tweet: “No Macedonian politician (from any political party) has the right to negotiate or change our name, identity and history. They attack and lie about those who defend our basic human rights but won’t defend themselves against those who admit to wanting to erase our identity.”

#Бојкотирам#Macedoniapic.twitter.com/zG1ojOBTD2

— Meto Koloski (@MetodijaKoloski) July 31, 2018


Meto Koloski, president of United Macedonian Diaspora, tweets his support for boycotting the Prespa referendum in July 2018. Experts say diaspora groups became active in drumming up support for the boycott campaign after getting fired up on social media.

Kostovski told BIRN the #boycott movement received several thousand euros from Todor Petrov, leader of the World Macedonian Congress, a Skopje-based non-governmental organisation that boasts diaspora members in the United States, Canada, Australia, Italy and Germany.

In 1991, Petrov had advocated putting the Vergina Sun on the new country’s flag. His World Macedonian Congress is widely seen as an ultranationalist movement.

“The World Macedonian Congress has connections with many Macedonians around the world and it is true that they helped campaign for last year’s referendum boycott,” Petrov to BIRN.

Meanwhile, Kostovski said other politicians “tried to ride the wave we created. And they all demanded – and got – a lot of money from the Macedonian diaspora.”

One such beneficiary was Janko Bacev, president of the pro-Russian United Macedonia party, he said — though BIRN was unable to confirm the claim.

Asked to comment, Bacev told BIRN: “I won’t comment on provocateurs working for the puppet government in Macedonia.”

Bacev was seen at a violent anti-Prespa protest in front of parliament in June 2018 that police quelled with teargas and stun grenades.

People from all over Greece protest against the Prespa agreement at a rally in Syntagma Square, Athens, on 20 January 2019. Photo: EPA-EFE/ORESTIS PANAGIOTOU

‘We want our name back’

BIRN’s investigation showed that Greek nationalists on the other side of the border also used computational propaganda to whip up a backlash against the Prespa deal.

Again, diaspora activists played a role in turning online propaganda into action on the street, with anti-Prespa anger fuelling the biggest protests in Athens and Thessaloniki since Greece’s debt crisis.

“This is a geographical area where large populations were forced in the 20th Century to migrate for economic and political reasons, first to the US, Canada and then to Australia,” said Tasos Kostopoulos, a historian and investigative journalist at the Efimerida ton Sintakton daily paper in Athens.

“It is precisely these people, especially the second and third generation, who are involved in a raging fight online, exchanging insults on Twitter with hundreds of trolls and bots.”

Like Cheese, one of the loudest Greek voices in the digital cacophony is a human-machine hybrid, according to BIRN’s investigation.

“I AM A GREEK MACEDONIAN!” says the Twitter profile of “Pallas Athena” under an image of the Vergina Sun. “We 3.5 million Greek Macedonians are tired of being robbed of our IDENTITY, HISTORY, NAME AND SYMBOLS! We want our name Macedonia back!”

The TweetBotorNot application rates the chance of ‘Pallas Athena’ being a bot at almost 70 per cent. Since the account clearly uses automation in addition to real human intervention, it is what is known in the online propaganda business as a ‘cyborg bot’. Image courtesy of Michael W. Kearney

The sheer number of tweets from Pallas Athena’s account – around 478,000 in five years – is a clear indication of automation, though plenty of the content is clearly human-generated too.

According to analysis by TweetBotOrNot, a software application that uses machine learning to classify Twitter accounts as bots or human, there is an almost 70 per cent chance that Pallas Athena is a cyborg bot.

Using advanced metrics and monitoring tools, BIRN extracted and analysed a week’s worth of output from the account – more than 2,500 tweets and retweets.

The number-crunching revealed that Pallas Athena’s online actions – human or otherwise – potentially reached no fewer than 9.7 million other Twitter users in seven days.

Geolocation analysis showed these users were in 106 spots across the globe including Athens, Toronto, Caracas, Miami and Melbourne.

Experts say such numbers show the power of computational propaganda to create an ever-expanding echo chamber from a single account.

The Twitterbots crawler application shows ‘Pallas Athena’ at the centre of a network of bots and real Twitter users — the ultimate digital echo chamber. Image: Twitterbots screengrab

Contacted by BIRN, the owner of the Pallas Athena account messaged: “Dear, I am not a bot.”

Dear, I am not a bot.

‘Pallas Athena’

She identified herself as a Macedonian woman living permanently in Sweden who took to Twitter in 2014 after she “saw the people from Skopje claiming that they have suffered genocide from the Greeks”.

“Unthinkable,” she wrote. “I come from Alexander the Great’s [ancient city of] Pella and always my grandmother, Helen, used to talk to me about the crimes of Bulgarians and Turks in the area.

“I started looking into old newspaper records and understood that they [people living in what is now North Macedonians] were committing the real crimes. So I started actively working on Twitter with the Macedonian issue.”

In April 2018, she locked horns with Cheese in a public Twitter spat.

I AM A GREEK MACEDONIAN FROM EDESSA PELLA IN REAL ANCIENT GREEK MACEDONIA , THE AREA WERE ALEXANDER THE GREAT WAS BORN !
Your country has bever been part of Ancient Greek Macedonia , it is in the 7
Region of Ancient Paeonia and you are not Macedonians ! pic.twitter.com/siXENjcijp

— Pallas Athena (@Makedni) April 15, 2018

“Good night, fellow Greek,” Cheese taunted her. “Good night from a Macedonian from Macedonia.”

Pallas Athena replied: “I AM A GREEK MACEDONIAN FROM EDESSA PELLA IN REAL ANCIENT GREEK MACEDONIA, THE AREA WHERE ALEXANDER THE GREAT WAS BORN! Your country has never been part of Ancient Greek Macedonia.”

After more exchanges like that, the two cyborg bots blocked each other on Twitter and got on with other business.

For Nikos Smyrnaios, a professor of political economy and the sociology of media and the internet at the University of Toulouse, blaming such animus on technology is only part of the story.

“It was not the technology that shaped this deep polarisation in the two countries but the very societies that for decades kept creating the conditions for this computational nationalist propaganda to grow and take root,” he said.

Kostas Zafeiropoulos is an investigative reporter for Efimerida ton Sintakton in Athens. This article was produced as part of the Balkan Fellowship for Journalistic Excellence, supported by the ERSTE Foundation, in cooperation with the Balkan Investigative Reporting Network. Editing by Timothy Large.

Amnesty International Updates Citizen Evidence Lab

Amnesty International has launched an updated version of its Citizen Evidence Lab website, bringing together cutting-edge open-source and other digital investigation tools which have revolutionized how evidence of serious human rights violations and other crimes are gathered and preserved.

Investigations facilitated by the pioneering Citizen Evidence Lab website have already helped expose human rights violations Cameroon, war crimes in Syria and chemical weapons attacks in Sudan.

The upgraded site provides a space for human rights researchers, investigators, students and journalists to explore and share investigative techniques in human rights. It enables them to take better advantage of the digital data-streams critical for modern fact-finding, while also leading the fight against mis- and disinformation campaigns.

“Human rights investigations in the digital age are constantly evolving, and the Citizen Evidence Lab was originally created as a space to keep on top of innovations by sharing tips, tools and best practices on disciplines such as video verification, remote sensing and weapons analysis,” said Sam Dubberley, acting head of the Crisis Response Programme’s Evidence Lab at Amnesty International.

 The Evidence Lab brings together investigators, engineers, developers and others to pilot new and expanding tools such as artificial intelligence, remote sensing, weapons identification and big-data analytics.

Evidence Lab initiatives feed into dozens of Amnesty International research reports, press releases and other outputs each year. It also creates large-scale, standalone collaborative projects involving volunteers around the world, including:

  • Amnesty Decoders: a crowd-source network of tens of thousands of activists to process large volumes of data such as satellite imagery, documents, pictures or social media messages. Decoders projects aim to go beyond “clicktivism,” enabling volunteers to generate meaningful data for Amnesty International’s human rights investigations.
  • The Digital Verification Corps (DVC), a network of more than 100 multidisciplinary students at six partner universities who authenticate videos and images found on social media to support human rights research in a more complicated world of mis- and dis-information. The programme recently won the prestigious 2019 Times Higher Education award for International Collaboration.

The Evidence Lab has contributed to high-profile, impactful human rights investigations, building on Amnesty International’s legacy of pioneering citizen evidence and remote sensing, dating back to the ground-breaking Eyes on Darfur project in 2007. Just a few examples include:

Freedom on the Net: Tracking the Global Decline

Freedom on the Net is a comprehensive study of internet freedom in 65 countries around the globe, covering 87 percent of the world’s internet users. It tracks improvements and declines in internet freedom conditions each year. The countries included in the study have been selected to represent diverse geographical regions and regime types. In-depth reports on each country can be found at freedomonthenet.org.

More than 70 analysts contributed to this year’s edition, using a 21-question research methodology that addresses internet access, freedom of expression, and privacy issues. In addition to ranking countries by their internet freedom score, the project offers a unique opportunity to identify global trends related to the impact of information and communication technologies on democracy. Country-specific data underpinning this year’s trends is available online. This report, the ninth in its series, focuses on developments that occurred between June 2018 and May 2019.

Of the 65 countries assessed, 33 have been on an overall decline since June 2018, compared with 16 that registered net improvements. The biggest score declines took place in Sudan and Kazakhstan followed by Brazil, Bangladesh, and Zimbabwe.

In Sudan, nationwide protests sparked by devastating economic hardship led to the ouster of President Omar al-Bashir after three decades in power. Authorities blocked social media platforms on several occasions during the crisis, including a two-month outage, in a desperate and ultimately ineffective attempt to control information flows. The suspension of the constitution and the declaration of a state of emergency further undermined free expression in the country. Harassment and violence against journalists, activists, and ordinary users escalated, generating multiple allegations of torture and other abuse.

In Kazakhstan, the unexpected resignation of longtime president Nursultan Nazarbayev—and the sham vote that confirmed his chosen successor in office—brought simmering domestic discontent to a boil. The government temporarily disrupted internet connectivity, blocked over a dozen local and international news websites, and restricted access to social media platforms in a bid to silence activists and curb digital mobilization. Also contributing to the country’s internet freedom decline were the government’s efforts to monopolize the mobile market and implement real-time electronic surveillance.

The victory of Jair Bolsonaro in Brazil’s October 2018 presidential election proved a watershed moment for digital election interference in the country. Unidentified actors mounted cyberattacks against journalists, government entities, and politically engaged users, even as social media manipulation reached new heights. Supporters of Bolsonaro and his far-right “Brazil over Everything, God above Everyone” coalition spread homophobic rumors, misleading news, and doctored images on YouTube and WhatsApp. Once in office, Bolsonaro hired communications consultants credited with spearheading the sophisticated disinformation campaign.

In Bangladesh, citizens organized mass protests calling for better road safety and other reforms, and a general election was marred by irregularities and violence. To maintain control over the population and limit the spread of unfavorable information, the government resorted to blocking independent news websites, restricting mobile networks, and arresting journalists and ordinary users alike.

Deteriorating economic conditions in Zimbabwe made the internet less affordable. As civil unrest spread throughout the country, triggering a violent crackdown by security forces, authorities restricted connectivity and blocked social media platforms.

China confirmed its status as the world’s worst abuser of internet freedom for the fourth consecutive year. Censorship reached unprecedented extremes as the government enhanced its information controls in advance of the 30th anniversary of the Tiananmen Square massacre and in the face of widespread antigovernment protests in Hong Kong. In a relatively new tactic, administrators shuttered individual accounts on the hugely popular WeChat social media platform for any sort of “deviant” behavior, including minor infractions such as commenting on environmental disasters, which encouraged pervasive self-censorship. Officials have reported removing tens of thousands of accounts for allegedly “harmful” content on a quarterly basis. The campaign cut individuals off from a multifaceted tool that has become essential to everyday life in China, used for purposes ranging from transportation to banking. This blunt penalty has also narrowed avenues for digital mobilization and further silenced online activism.

Internet freedom declined in the United States. While the online environment remains vibrant, diverse, and free from state censorship, this report’s coverage period saw the third straight year of decline. Law enforcement and immigration agencies expanded their surveillance of the public, eschewing oversight, transparency, and accountability mechanisms that might restrain their actions. Officials increasingly monitored social media platforms and conducted warrantless searches of travelers’ electronic devices to glean information about constitutionally protected activities such as peaceful protests and critical reporting. Disinformation was again prevalent around major political events like the November 2018 midterm elections and congressional confirmation hearings for Supreme Court nominee Brett Kavanaugh. Both domestic and foreign actors manipulated content for political purposes, undermining the democratic process and stoking divisions in American society. In a positive development for privacy rights, the Supreme Court ruled that warrants are required for law enforcement agencies to access subscriber-location records from third parties.

Only 16 countries earned improvements in their internet freedom scores, and most gains were marginal. Ethiopia recorded the biggest improvement this year. The April 2018 appointment of Prime Minister Abiy Ahmed led to an ambitious reform agenda that loosened restrictions on the internet. Abiy’s government unblocked 260 websites, including many known to report on critical political issues. Authorities also lifted a state of emergency imposed by the previous government, which eased legal restrictions on free expression, and reduced the number of people imprisoned for online activity. Although the government continued to impose network shutdowns, they were temporary and localized, unlike the nationwide shutdowns that had occurred in the past.

Other countries also benefited from an opening of the online environment following political transitions. A new coalition government in Malaysia made good on some of its democratic promises after winning May 2018 elections and ending the six-decade reign of the incumbent coalition. Local and international websites that were critical of the previous government were unblocked, while disinformation and the impact of paid commentators known as “cybertroopers” began to abate. However, these positive developments were threatened by a rise in harassment, notably against LGBT+ users and an independent news website, and by the 10-year prison term imposed on a user for Facebook comments that were deemed insulting to Islam and the prophet Muhammad.

In Armenia, positive changes unleashed by the 2018 Velvet Revolution continued, with reformist prime minister Nikol Pashinyan presiding over a reduction in restrictions on content and violations of users’ rights. In particular, violence against online journalists declined, and the digital news media enjoyed greater freedom from economic and political pressures.

Iceland became the world’s best protector of internet freedom, having registered no civil or criminal cases against users for online expression during the coverage period. The country boasts enviable conditions, including near-universal connectivity, limited restrictions on content, and strong protections for users’ rights. However, a sophisticated nationwide phishing scheme challenged this free environment and its cybersecurity infrastructure in 2018.

BIRD Community

Are you a professional journalist or a media worker looking for an easily searchable and comprehensive database and interested in safely (re)connecting with more than thousands of colleagues from Southeastern and Central Europe?

We created BIRD Community, a place where you can have it all!

Join Now