Google says mobile-first indexing is now used for over half the web pages in its search results

Google announced today it’s now using mobile-first indexing for over half the web pages shown in its search results globally – a significant milestone in Google’s move to favor mobile sites over desktop sites in its search results.

The plans for the project have been in the works for years.

The company had first detailed its efforts around mobile-first indexing back in 2016, where it explained the impacts to how its search index operates. It said it would shift over to using the mobile version of a website’s content to index its pages, as well as to understand its structured data and show snippets from the site in Google’s search results.

Its reasoning behind the change is simple: most people today search Google from a mobile device, not a desktop computer. But Google’s ranking systems for the web were originally built for the desktop era. They still typically look at the desktop version of the page’s content to determine its relevance to the user.

This, obviously, causes problems when the desktop site and the mobile site are not in sync.

Before responsive web design became more commonplace, many site owners built a separate, simpler and sometimes less informative version of their site for their mobile web visitors. These users may have been guided to the site because of Google Search. But once there, they couldn’t find what they were looking for because it was only available on the desktop version of the web page.

In December 2017, Google said it had begun to transition a small handful of sites to mobile-first indexing.

Earlier this year, Google announced it had begun to officially roll out its “mobile-first” indexing of the web, following a year and a half of testing and experimentation. At the time, it said it would first move over the sites already following the best practices for mobile-first indexing. It also noted it would favor the site’s own mobile version of its webpage over Google’s fast-loading AMP pages.

Sites who are shifted are notified through a message in Search Console and then see increased visits from the smartphone version of Googlebot, which crawls the mobile version of their site. Site owners can also check their server logs, where they can track the increased requests from Googlebot Smartphone.

Google additionally offers a URL inspection tool, which site owners can use to check how a URL from their site – like the homepage – was last crawled and indexed.

Google today notes that sites that don’t use responsive web design are seeing two common problems when Google tries to move them over to mobile-first indexing.

Some don’t use structured data on their mobile sites, even though they use it on the desktop. This is important because it helps Google to understand the website’s content and allows it to highlight pages’ content in its search results, through its “fancier” features like rich results, Knowledge Graph results, enhanced search results, carousels and more – basically any time you see more engaging search results that offer more than just a list of blue links.

The company also said that some mobile sites were missing alt-text for images, which makes it harder for Google to understand the images’ content.

At the time of the initial wave of sites being shifted over, Google had said that the mobile-friendly index wouldn’t directly impact how content is ranked, but it did say that a site’s mobile-friendly content will help it “perform better” in mobile search results. Mobile-friendliness has also long been one of many factors in determining how a site is ranked, but it’s not the only one.

Google didn’t say what it will do to sites that are never properly updated for the mobile web, but it seems that – at some point – their ranking could be impacted.

 

Kahoot, a ‘Netflix for education’, launches an accelerator to tap gaming and education startups

On the back of Disney increasing its shareholding in Oslo-based Kahoot to four percent last week, Kahoot today announced a new initiative that helps to position the popular startup — which already has 60 million games and has seen over 1 billion players engage on its platform over the last year — as the “Netflix for education apps.”

It’s launching Kahoot! Ignite, a new accelerator for like-minded startups that are pushing the boundaries of education through gaming and other means.

In addition to that, Kahoot today also said it would move stock exchanges in its home market of Norway, going from the smaller OTC exchange to the Merkur Market, which CEO and co-founder Åsmund Furuseth explained in an interview is also an exchange for private companies, but one that will be able to provide more transparency to the startup’s bigger investors en route to an eventual full public listing. As of last week’s Disney news, the startup is now valued at $376 million.

Participating in the Ignite accelerator, Furuseth said, will give Kahoot the option to invest in startups in each cohort, and if it makes sense for the startup in question, they will build content that will be usable on the Kahoot platform.

“We have close to $30 million in the bank and are in a financial market where we can get more capital,” he said. “We don’t need to invest, but if we want to, we can.” 

The startup today has some 60 million games on its platform, with a good portion of those created by users themselves (making it more like a YouTube than a Netflix). The idea is that bringing in outside developers (in this case, by way of the accelerator) could inject more innovation and interesting takes on the concept of “educational gaming” — not unlike how Netflix and Amazon engage outside studios to develop originals for its platform, alongside what they develop themselves or buy in through deals with rights holders.

In addition to the carrots of investment and distribution on the Kahoot platform — which is likely to hit 100 million monthly active users this month (Furuseth said he was confident of the number today) — Kahoot is offering mentorship to potential cohorts in areas like monetization and product development. Given the fact that educational aides can come in all shapes and sizes, that might not take the form of a piece of content for the Kahoot platform.

“Putting something on Kahoot could be an outcome, but we’re also interested in ‘network products,’ which have the same desire to enable learning,” Furuseth said.

The company today has a double focus, with games for K-12 students as well as for enterprise environments. “Learning is the main topic,” he added. “We like to have the mix.”

NHS nurse, 42, charged with FAKING attack on herself

Surprise! Michelle Obama and Jimmy Fallon stun tour groups in 30 Rock's elevators as former First Lady jokes that protective Barack only wanted a second term to ensure Malia and Sacha continued to get Secret Service protection Facebook 'secretly allowed Netflix and Spotify to read users' private messages and shared YOUR personal data with Silicon Valley giants Amazon and Microsoft as recently as last summer', as Zuckerberg is hit with yet ANOTHER privacy scandal Stormy weather to blanket the Eastern half of the U.S. and Pacific Northwest this weekend and cause Christmas travel chaos for 110MILLION Americans Trump's promise to 'review' case of the Green Beret murder suspect BEFORE it's gone to trial 'leaves military legal experts FURIOUS with the President's unwanted intervention' George Lucas tops Forbes list of America's wealthiest celebrities with estimated net worth of $5.6billion ... (more)

Price f(x) picks up €25M Series B for its pricing optimization SaaS

Price f(x), a startup that offers cloud-based pricing software, has raised €25 million in Series B funding. Leading the round is European B2B technology growth investor Digital + Partners, and consulting firm Bain & Company. Prague-based Credo Ventures and London-based Talis Capital, which both backed Price f(x) at Series A, have also participated in this new round.

Founded in 2011, Munich-based Price f(x) provides a modular SaaS solution for price optimisation management (PO&M) and configure-price-quote (CPQ) for enterprises of any size.

Pricing optimisation software typically helps companies accurately define the price of goods across a vast and constantly changing spectrum of data and variables. This can include things like customer survey data and segments, competitor data, operating costs, inventories, and historic prices and sales.

CPQ software aggregates these variables, thus enabling companies to configure products or services in the most optimal way (i.e. bundling, up-sells, etc.), and price them according to costs, competition and local economic factors.

This end result is that is that it can drastically speed up and improve the accuracy of the quoting process to give customers the best price possible in accordance with all of the above factors.

Price f(x) says it currently serves over 80 global, blue-chip B2B and B2C customers across a variety of industries, including Robert Bosch, SchneiderElectric, Owens- Illinois, Iron Mountain and Sonoco. The company has also developed a strong partner ecosystem with leading global technology, consulting and integration providers, including new backer Bain & Company, and SAP.

Notably, Price f(x) is in the midst of litigation with U.S. competitor Vendavo over a number of patent disputes. In December 2017, Vendavo launched a lawsuit against Price f(x), which the German company refutes. And earlier this week, Price f(x) filed petitions for “Covered Business Method (“CBM”) Review” of four Vendavo patents, and says it will imminently file a fifth, which together will cover challenges to all of the patent claims that Vendavo has asserted in litigation between the two parties.

“Price f(x) has become the leading SaaS pricing solution provider on the market through its customer centric approach and by offering a feature rich, highly flexible pricing tool that is also risk free and fast to implement,” says Marcin Cichon, CEO and co-Founder of Price f(x), in a statement. “Our success is based on the continued satisfaction and loyalty of our customers. This new funding will allow us to help even more businesses to thrive by further expanding our existing platform capabilities and also introducing a new product offering for the SME market segment”.

“For most companies, pricing is the single most effective lever to boost earnings,” says Ron Kermisch, Bain & Company’s global pricing leader. “Yet many companies leave money on the table because they do not set the best price or ensure customers actually pay the price they have determined. Bain & Company sees investing in Price f(x) as a great opportunity to help Price f(x) to become the de-facto standard in pricing and with that to be also the best-of-breed competitive weapon for Bain’s clients, to stay at the cutting edge of pricing”.

Flux raises $7.5M Series A to bring its digital receipts platform to more banks and merchants

Flux, the London fintech that has built a technology platform for banks and merchants to power itemised digital receipts and a lot more, has raised $7.5 million in Series A funding. The round is led by VC firm e.ventures (which has previously backed the likes of Farfetch, Sonos and Groupon), with participation from existing investors PROfounders, and Anthemis.

Founded in 2016 by former early employees at Revolut, Flux bridges the gap between the itemised receipt data captured by a merchant’s point-of-sale (POS) system and what little information typically shows up on your bank statement or mobile banking app. Off the back of this, it can also power loyalty schemes and card-linked offers, as well as give merchants much deeper POS analytics via aggregated and anonymised data on consumer behaviour, such as which products are selling best in unique baskets.

On the banking side, Flux is currently available through Barclays (via Barclays Launchpad), challenger bank Starling, and for a small alpha group of Monzo customers. Once banking customers link their account to the service, Flux delivers digital receipts (and where available rewards and loyalty) for transactions at Flux retailer partners.

To that end, merchant partnerships include Costa Coffee, EAT, pod and itsu. Flux also recently announced that Pure is joining the service.

“Our mission has always been to liberate the world’s receipt data because by doing this we can enrich trillions of experiences globally,” Flux co-founder and CEO Matty Cusden-Ross tells me.

“The information on a receipt is used all the time in everyday life, from budgeting to loyalty to expensing but today these all require manual steps. We see a future in which all of these manual processes become seamless experiences, simplifying and enriching people’s lives. Our focus today is on establishing a standard, the Flux platform, to make this a reality within the U.K. before expanding to our first international market”.

Of course, Flux’s attempts to become a standard for the interchange of item level digital receipt data — and the proprietary platform that powers that standard — has always faced a chicken and egg problem: It needs bank integrations to sign up merchants and it needs merchant integrations to sign up banks. Cracking this problem has clearly started to gather momentum, something that hasn’t gone unnoticed by investors.

“We’ve transitioned from having to prove it’s possible to now scaling and that’s a great feeling,” says Cusden-Ross. “The aim for this round is to continue making Flux the gold standard for anything that touches receipt data, [ensuring] Flux remains super easy to use for everyone — consumers, banks and retailers. What this means is going fully live across some of the largest U.K. retail banks as well as ramping our up our live merchants”.

(Related, I understand that Flux has already begun integrating with one of the major U.K. supermarkets and an “international fast food chain,” amongst other unannounced partnerships.)

“Creating a real-time platform that handles massive data volumes is hard, but we’ve cracked it,” adds the Flux CEO. “We’re investing heavily in bringing on the best engineers to continue scaling in a big way. Having figured out the recipe for working with banks and retailers quickly, it’s now all about growing as fast as possible”.

Zwift, which turns indoor cycling workouts into multiplayer games, raises $120M

Fitness and gaming have been two of the most popular categories of apps for years, and now a startup founded out of London that has combined the two in a unique way has picked up a big round of funding to capitalise on that. Zwift, an interactive platform for people to turn indoor cycling workouts into massive, multi-participant races, social rides, and immersive explorations of new domains, has raised $120 million — money that its co-founder and CEO, Eric Min, said will be used to expand to more training categories (it’s first steps outside of cycling have been into running), and to add esports tournaments.

The funding — led by Highland Europe, with Europe’s True (not to be confused with True Ventures), Causeway Media and Novator participating — comes on the heels of very rapid growth for Zwift .

The startup now has over 1 million registered accounts (it doesn’t disclose active users), up from a mere 200,000 two years ago, with users ranging from amateur cycling enthusiasts, people who cycle as part of fitness regimes, and professional athletes who use it to supplement IRL global training schedules.

“More than one-third of the peloton in this year’s Tour de France” — despite all its controversies, still the gold standard for road races — “are users of Zwift,” Min told me. There are 200 Facebook groups built around Zwift communities, with people using them to organise rides and sometimes even to meet up in person afterwards at cafes, just as they might in actual outdoor rides. With customers in 100 countries, Min told me that there are on average 300 rides a day happening on Zwift.

The company isn’t disclosing its valuation, but Min said that the startup is “approaching unicorn status” on account of its growth and big ambitions and it appears to be more than $500 million with “a very small dilution” — a big jump on the $180 million Pitchbook estimated for its valuation in 2016 (it’s raised around $166 million to date).

Min believes that the upswing in e-sports could see the format getting ever-more mainstream acceptance. “Our goal is to bring Zwift to the Olympics,” he told me confidently.

Zwift is swiftly addressing two different shortcomings of two common home pastimes. One of the problems with exercising at home — and specifically training in cycling as a first effort — is that it might get a little boring, and one of the problems with playing too many video games is that they contribute to the tendency we have in our modern world to be too sedentary.

The service involves you providing your own bike, which you link up with a Zwift trainer (a rack-like piece of equipment that turns a bike into a stationary bike for indoor training), which in turn picks up your stats and adjusts tension and so on based on the course that you are riding. You cycle in front of a TV typically to get the immersive effect, linked up to a Mac, Windows or iOS App (it’s also on Apple TV). You start with a free trial before moving to a monthly fee of $15 (or $10 if you are currently on a trial or already subscribe: the higher fee was introduced last month).

There are no plans at the moment for VR headsets or other head-based wearables because so far they have proven to be too bulky to be usable in the physical environment of sometimes-gruelling cycling, Min said. And for now, you also don’t use spinners or other stationary cycling apparatuses because these can’t provide the right kind of ‘real-world’ riding feel, he said. But this might evolve as Zwift partners with more third parties (and with companies like Peloton a big hit with home fitness enthusiasts, you can see how that might evolve).

In all of its sports and sport ambitions, Zwift is bringing not just basic tension and gaming dynamics to the table: it’s using some interesting algorithms to help train its users, and figure out what might be the best logical step for them in terms of increasing or decreasing difficulty while measuring all other cycling stats. Min says this all starts with getting an accurate weight for each user.

Cycling accounts for 98 percent of the company’s business at the moment, and running is just taking off. For runners, people use running treadmills, and that is the template the company wants to follow, eventually including rowing machines, step machines, and more. Sometimes the locations are actual places like Innsbruck, pictured above. Sometimes they are make-believe terrains, like the fictional Pacific Island Watopia, below:

Min tells me he started Zwift out of London because that is where he lives, but much of the development has come out of southern California, since that is where there is a stronghold of gaming developers. Min himself is an expat from America (there are a lot of us here!), New York specifically, who previously had been a VP at JP Morgan and then founded a fintech company called Sakkonet Technology that’s still going strong, and he happens to be a cycling enthusiast going back to childhood.

“I wanted to do something different from fintech and the idea for Zwift was starting right at me,” he said. “I was a cyclist since I was a kid, painfully so. The worst thing you could do is ride your bike indoors because it was so boring. I used all the products I could but it wasn’t engaging enough, so we thought, what about taking the tech that out there for gaming, combining that with what you do in cycle training, and bring all that to a digital platform? Even if you could get 80 percent of the experience, that would be better than what was offered in the past.” 

Highland and the other investors are all strategic, in that they are already investing quite a bit in other fitness enterprises, which could in turn point to partnerships down the road. Highland’s in eGym and Huel; True backs the Ribble cycling brand; and Causeway invests in traditional sports and esports.

“Zwift is a fantastically innovative company and they are certainly leading the way in the indoor training space,” says Tony Zappala, Partner of Highland Europe in a statement. “It’s a highly scalable business and we’ve been impressed with how they have already managed to expand globally – already 70 percent of current subscribers are from outside the USA. Research points to an audience of 40 million competitive and enthusiast cyclists, and many of those lie in the traditional cycling nations of Central Europe, so in this market alone there is huge growth potential.”

 

 

SoftBank Corp shares drop 14% on their first day of trading, but it’s still one of the largest IPOs ever

SoftBank Corp’s initial public offering today started with a bang before trailing off into a whimper, with the stock falling 14.5 percent during its first day of trading on the Tokyo Stock Exchange.

The company is the mobile unit of conglomerate SoftBank Group, whose holdings also include Sprint and the $100 billion Vision Fund.

Shares of SoftBank Corp opened at 1,463 yen, below the 1,500 yen the company had set for its IPO price (instead of a range), and closed at 1,282 yen. It offered 160 million shares, or about a third of the total held by parent company SoftBank Group. Despite a bumpy first day of trading, SoftBank Corp raised a total of 2.65 trillion yen (about $23.5 billion), making it Japan’s largest ever IPO and placing it just behind Alibaba’s record-setting $25 billion debut on the New York Stock Exchange in 2014 (SoftBank Group is one of Alibaba’s largest shareholders).

According to Bloomberg, 90 percent of the investors who bought SoftBank Corp shares at the 1,500 opening price were individuals, who the company had targeted in an unusual marketing campaign.

Factors that may have dampened investor enthusiasm about include a network outage earlier this month triggered by a shutdown of Ericsson equipment due to expired software certificates (O2 customers in Great Britain were also affected).

The outage underscored other concerns about SoftBank Corp’s telecommunications infrastructure. According to a Nikkei report published last week, the company has decided to stop using hardware from Huawei Technologies due to security concerns and replace them over the next several years with equipment by Ericsson and Nokia.

While the company says the hardware swap isn’t expected to cost a lot of money, it will also need to deal with more competition next year. SoftBank Corp’s rivals are currently NTT DoCoMo and KDDI, but Rakuten will launch cellular service in October 2019, making it Japan’s fourth mobile network operator.

Furthermore, SoftBank Group also carries massive debt that totaled 18 trillion yen (about $160 billion) as of the end of September, more than six times the amount it earns on an operating basis. This means the Vision Fund is especially reliant on Saudi Arabia’s sovereign fund, which contributed $48 billion, making it the fund’s largest investor.

Saudi Arabia’s sovereign fund, called the Public Investment Fund, is run by Saudi Crown Prince Mohammed bin Salman, who has been implicated by Turkish officials and the United State’s Central Intelligence Agency in the planning of journalist Jamal Khashoggi’s murder. Crown Prince bin Salman has denied involvement in the killing, but the situation still calls into question the future of Saudi Arabia’s involvement with SoftBank, especially since Crown Prince bin Salman said in October that Saudi Arabia plans to invest another $45 billion in the second Vision Fund.

Facebook purges more ‘bad actors’ in Myanmar but it still won’t commit to a local office

As Facebook continues to grasp the severity of the situation in Myanmar, where the UN has concluded that its social network plays “determining role” in inciting genocide, the U.S. tech giant has completed a third sweep in recent months to remove bad actors from its platform.

Facebook said late Tuesday U.S. time that it has removed a total of 135 Facebook accounts, 425 Pages, 17 Groups and an additional 15 Instagram accounts with this latest piece of action.

Facebook has around 20 million users in Myanmar — that’s nearly all of the country’s internet users and nearly 40 percent of the population — and it gave some stats on the reach that it has now nullified:

  • Approximately 2.5 million people followed at least one of these Facebook Pages
  • Approximately 6,400 people belonged to at least one of these Facebook Groups
  • Approximately 1,300 people followed at least one these Instagram accounts

This is Facebook’s third such cull in recent months. Its previous removals impacted some high-profile individuals including Senior General Min Aung Hlaing, commander-in-chief of the armed forces, and the military-owned Myawady television network were removed from the social network following “evidence [that they] committed or enabled serious human rights abuses in the country.”

What’s notable about this newest action is that the company said it took action because of “the behavior of these actors rather than on the type of content they were posting.”

We’re waiting for further confirmation on exactly what that means, but acting irrespective of posted content would represent an interesting change in its policing, and it could impact Facebook’s efforts in Myanmar — and other areas — going forward.

Nearly everyone who has internet access in Myanmar uses Facebook, giving it an estimated user base of around 20 million. AFP PHOTO / Nicolas ASFOURI / Getty Images

That’s promising but, unfortunately, it appears that Facebook is still reluctant to commit to opening a local office in Myanmar. That’s something that local civic groups on the ground in Myanmar — who have worked with Facebook to improve the situation — have called a key requirement for meaningful progress.

“How many companies have 20 million users in one country but don’t have a single employee, it’s absurd,” Jes Petersen — CEO of accelerator firm Phandeeyar, which is part of the advisory group — told TechCrunch last month. “An office would go a long way to building relationships with stakeholders.”

Facebook declined to comment on the possibility of a Myanmar-based office when we asked.

The company has pledged to increase the number of Burmese translators working on Myanmar-based content to 100 by the end of this year. It has said a number of times that it is working on AI-based solutions, too, but cracks still appear.

Equally, while reaching 100 translators means Facebook has more than doubled its Burmese-compliant content checking contingent, the figure is dwarfed by others. Myanmar’s army reportedly has 700 people working on its own Facebook strategy.

Sources familiar with the company’s thinking told TechCrunch that Facebook is concerned that “there would be real risks involved” if it were to open an office, “including the potential for increased government leverage on content and data requests as well as potential risks to Facebook’s employees.”

That response is backed, according to the sources, by the findings of a BSR report that was released last month.

If this is consistent with the company’s strategy then it is troubling because that doesn’t tell the whole truth of what is a very nuanced issue.

While it is correct that the report did mention the potential risks associated with an office — around both the safety of staff and potential for government pressure — the conclusion wasn’t that Facebook shouldn’t open the office. It was that there are “advantages and disadvantages” to it doing so.

So you could equally argue that it should open an office if you choose to focus the positive argument from the report.

More generally, it is certainly ironic that Facebook is (partially) citing insight from a report that it controversially released on the eve of the U.S. mid-term elections, a move that many took as an effort to bury the findings while the news cycle was focused on a key political moment.

While it may not get the same press attention as Russian-backed U.S. election meddling, the Facebook-Myanmar situation is a key one to watch in 2019. Facebook is the de facto internet in Southeast Asia and other emerging markets so its influence extends beyond anything people in Western markets can begin to imagine.

Amnesty International used machine-learning to quantify the scale of abuse against women on Twitter

A new study by Amnesty International and Element AI puts number to a problem many women already know about: that Twitter is a cesspool of harassment and abuse. Conducted with the help of 6,500 volunteers, the study, billed by Amnesty International as “the largest ever” into online abuse against women, used machine-learning software from Element AI to analyze tweets sent to a sample of 778 women politicians and journalists during 2017. It found that 7.1%, or 1.1 million, of those tweets were either “problematic” or “abusive,” which Amnesty International said amounts to one abusive tweet sent every 30 seconds.

On an interactive website breaking down the study’s methodology and results, Amnesty International said many women either censor what they post, limit their interactions on Twitter, or just quit the platform altogether. “At a watershed moment when women around the world are using their collective power to amplify their voices through social media platforms, Twitter’s failure to consistently and transparently enforce its own community standards to tackle violence and abuse means that women are being pushed backwards towards a culture of silence,” stated the human rights advocacy organization.

Amnesty International, which has been researching abuse against women on Twitter for the past two years, signed up 6,500 volunteers for what it refers to as the “Troll Patrol” after releasing another study in March 2018 that described Twitter as a “toxic” place for women. The Troll Patrol’s volunteers, who come from 150 countries and range in age from 18 to 70 years old, received training about constitutes a problematic or abusive tweet. Then they were shown anonymized tweets mentioning one of the 778 women and asked whether or not the tweets were problematic or abusive. Each tweet was shown to several volunteers. In addition, Amnesty International said “three experts on violence and abuse against women” also categorized a sample of 1,000 tweets to “ensure we were able to assess the quality of the tweets labelled by our digital volunteers.”

The study defined “problematic” as tweets “that contain hurtful or hostile content, especially if repeated to an individual on multiple occasions, but do not necessarily meet the threshold of abuse,” while “abusive” meant tweets “that violate Twitter’s own rules and include content that promote violence against or threats of people based on their race, ethnicity, national origin, sexual orientation, gender, gender identity, religious affiliation, age, disability, or serious disease.”

In total, the volunteers analyzed 288,000 tweets sent between January to December 2017 to the 778 women studied, who included politicians and journalists across the political spectrum from the United Kingdom and United States. Politicians included members of the U.K. Parliament and the U.S. Congress, while journalists represented a diverse group of publications including The Daily Mail, The New York Times, Guardian, The Sun, gal-dem, Pink News, and Breitbart.

Then a subset of the labelled tweets was processed using Element AI’s machine-learning software to extrapolate the analysis to the total of 14.5 million tweets that mentioned the 778 women during 2017. (Since tweets weren’t collected for the study until March 2018, Amnesty International notes that the scale of abuse was likely even higher because some abusive tweets may have been deleted or made by accounts that were suspended or disabled). Element AI’s extrapolation produced the finding that 7.1% of tweets sent to the women were problematic or abusive, amounting to 1.1 million tweets 2017.

Black, Asian, Latinx, and mixed race women were 34% more likely to be mentioned in problematic or abusive tweets than white women. Black women in particular were especially vulnerable: they were 84% more likely than white women to be mentioned in problematic or abusive tweets. One in 10 tweets mentioning black women in the study sample was problematic or abusive, compared to one in 15 for white women.

“We found that, although abuse is targeted at women across the political spectrum, women of color were much more likely to be impacted, and black women are disproportionately targeted. Twitter’s failure to crack down on this problem means it is contributing to the silencing of already marginalized voices,” said Milena Marin, Amnesty International’s senior advisor for tactical research, in the statement.

Breaking down the results by profession, the study found that 7% of tweets that mentioned the 454 journalists in the study were either problematic or abusive. The 324 politicians surveyed were targeted at a similar rate, with 7.12% of tweets that mentioned them problematic or abusive.

Of course, findings from a sample of 778 journalists and politicians in the U.K. and U.S. is difficult to extrapolate to other professions, countries, or the general population. The study’s findings are important, however, because many politicians and journalists need to use social media in order to do their jobs effectively. Women, and especially women of color, are underrepresented in both professions, and many stay on Twitter simply to make a statement about visibility, even though it means dealing with constant harassment and abuse. Furthermore, Twitter’s API changes means many third-party anti-bullying tools no longer work, as technology journalist Sarah Jeong noted on her own Twitter profile, and the platform has yet to come up with tools that replicate their functionality.

Amnesty International’s other research about abusive behavior towards women on Twitter includes a 2017 online poll of women in 8 countries, and an analysis of abuse faced by female members of Parliament before the UK’s 2017 snap election. The organization said the Troll Patrol isn’t about “policing Twitter or forcing it to remove content.” Instead, the organization wants the platform to be more transparent, especially about how the machine-learning algorithms it uses to detect abuse.

Because the largest social media platforms now rely on machine learning to scale their anti-abuse monitoring, Element AI also used the study’s data to develop a machine-learning model that automatically detects abusive tweets. For the next three weeks, the model will be available to test on Amnesty International’s website in order to “demonstrate the potential and current limitations of AI technology.” These limitations mean social media platforms need to fine-tune their algorithms very carefully in order to detect abusive content without also flagging legitimate speech.

“These trade-offs are value-based judgements with serious implications for freedom of expression and other human rights online,” the organization said, adding that “as it stands, automation may have a useful role to play in assessing trends or flagging content for human review, but it should, at best, be used to assist trained moderators, and certainly should not replace them.”

TechCrunch has contacted Twitter for comment.

Facebook’s got 99 problems but Trump’s latest “bias” tweet ain’t one

By any measure Facebook hasn’t had the best of years in 2018.

But while toxic problems keep piling up and, well, raining acidly down on the social networking giant — from election interference, to fake accounts, faulty metrics, security flaws, ethics failuresprivacy outrages and much more besides — the silver lining of having a core business now widely perceived as hostile to democratic processes and civilized sentiment, and the tool of choice for shitposters agitating for hate and societal division, well, everywhere in the world, is that Facebook has frankly far more important things to worry about than the latest anti-tech-industry salvo from President Trump.

In an early morning tweet today, Trump (again) attacked what he dubbed anti-conservative “bias” in the digital social sphere — hitting out at not just Facebook but tech’s holy trinity of social giants, with a claim that “Facebook, Twitter and Google are so biased towards the Dems it is ridiculous!”

Time was when Facebook was so sensitive to accusations of internal anti-conservative bias that it fired a bunch of journalists it had contracted and replaced them with algorithms — which almost immediately pumped up a bunch of fake news. RIP irony.

Not today, though.

When asked if it had a response to Trump’s accusation of bias a Facebook spokesperson told us: “We don’t have anything to add here.”

The brevity and alacrity of the response suggested the spokesperson had a really cheerful expression on their face when they typed it.

The relief of Facebook not having to give a shit this time was kinda palpable, even in pixel form.

It was also a far cry from the screeds the company routinely dispenses these days to try to muffle journalistic — and indeed political — enquiry.

Trump evidently doesn’t factor ‘bigly’ on Facebook’s oversubscribed risk-list.

Even though Facebook was the first name on the president’s (non-alphabetical) tech giant hit-list.

Still, Twitter appeared to have irked Trump more, as his tweet singled out the short-form platform — with an accusation that Twitter has made it “much more difficult for people to join [sic] @realDonaldTrump”. (We think by “join” he means follow. But we’re speculating wildly.)

This is perhaps why Twitter felt moved to provide a response to the claim of bias, albeit also without wasting a lot of words.

Here’s its statement:

Our focus is on the health of the service, and that includes work to remove fake accounts to prevent malicious behavior. Many prominent accounts have seen follower counts drop, but the result is higher confidence that the followers they have are real, engaged people.

Presumably the president failed to read our report, from July, when we trailed Twitter’s forthcoming spam purge, warning it would result in users with lots of followers taking a noticeable hit in the coming days. In a word: Sad.

Of course we also asked Google for a response to Trump’s bias claim. But just got radio silence.

In similar “bias” tweets from August the company got a bigger Trump-lashing. And in a response statement then it told us: “We never rank search results to manipulate political sentiment.”

Google CEO Sundar Pichai has also just had to sit through some three hours of questions from Republicans in Congress on this very theme.

So the company probably feels it’s exhausted the political bias canard.

Even while, as the claims drone on and on, it might truly come to understand what it feels like to be stuck inside a filter bubble.

In any case there are far more pressing things to accuse Google’s algorithms of than being ‘anti-Trump’.

So it’s just as well it didn’t waste time on another presidential sideshow intended to distract from problems of Trump’s own making.