What is all the fuss about Cambridge Analytica? Part 3

So Part 1 described the background and origins of Cambridge Analytica (CA).

Part 2 described the objective of Robert Mercer and with Carole Cadwallar describing the impact on Vote Leave.

This part is about the methods and implications of CA on electoral processes.

Hitting the Headlines

In March 2018, multiple media outlets broke news of Cambridge Analytica’s business practices: The New York Times and The Observer reported that the company had used Facebook data for its campaign activities and shortly afterwards, Channel 4 News aired undercover investigative videos showing CA CEO Alexander Nix boasting about using prostitutes, bribery sting operations, and honey traps to discredit politicians on whom it conducted “opposition research”. CA claimed it had “ran all of (Donald Trump’s) digital campaign” in 2016 Presidential election. In response in the UK, the Information Commissioner’s Office (ICO) issued a warrant to search the company’s servers. Meanwhile Facebook banned CA from advertising on its platform, saying that it had been deceived. On 23 March 2018, the ICO was granted a warrant to search Cambridge Analytica’s London offices.

Amazon said that they suspended CA from using their Cloud Hosting Services. The governments of India and Brazil demanded that CA report how their data was used in political campaigning.

In early July 2018, the United Kingdom’s Information Commissioner’s Office announced it intended to fine Facebook £500,000 ($663,000) over the data scandal, this being the maximum fine allowed at the time of the breach, saying Facebook “contravened the law by failing to safeguard people’s information”.

Also in July 2019, in the USA, Facebook was fined $5billion for its minor part in the data breach.

Rather too late… CA is now “gone” but their methods, their “genie” is now out of the bottle.

What did they do?

Wikipedia describes the method CA used to gain personal data. CA developed a Facebook app called “This Is Your Digital Life.” Aleksandr Kogan, a data scientist at Cambridge University, developed the app sometimes called “thisisyourdigitallife” and provided the app to CA who then posted it to Facebook. This third-party app then had permission to acquire data from Facebook users that not only entered data into a quiz-like game but also gave the app access to information on the user’s friends network; this resulted in the data of about 87 million users, the majority of whom had no idea their personal data was being collected for political ends. It goes without saying that the app breached Facebook’s terms of service but Facebook did not police any app particularly well (hence the reason for the $5b fine).

Follow this link to hear Alexander Nix describe the CA Big Data approach elections or this one to hear how big data helped Senator Ted Cruz in 2016. Nix claims that CA had 4000 parameters for every voter in the USA. From these parameters, not only demographics and location were uncovered but also psychographic profiles, the attitudes of each person distilled down to a few variables! This allowed, for any given political campaign, what kind of advertisement would be most effective to persuade a particular person for vote (or not vote) for any particular candidate or cause.

What CA has invented is the technology to subvert the traditional election processes to introduce:

  • Personalised messages – Nix claims top down broadcasting is dead. All future elections will be personalised messages based on a person’s psychographic profile.
  • Psychographic profiles are used to identify, and then reinforce, bias and prejudices.
  • Political promises are not on mainstream media, so not open to secutiny and debate, but are on social media. Fired up and forgotten with no follow up – reverting back to before Hansard when politicians were not held to account for any commitments.
  • Complete Situation Awareness of each individual’s motivations so that in all probability, each person can be manipulated using targeted messages to vote in the way expected (plus leaving the election to nefarious manipulation).

What are Psychographics?

Psychographic profiles can be valuable in the fields of marketing, demographics, opinion research, prediction, and social research in general.

All the research for political ends has already been established for marketing and advertising of products. Demographic information includes gender, age, income, marital status – the dry facts. In the past marketing was all about Demographics: making sure your advert went out to males or females of a partical age. Psychographics are kind of like demographics. Psychographic information might be your buyer’s habits, hobbies, spending habits and values. Demographics explain “who” the buyer is, while psychographics explain “why” they buy. Advertisers now reach their target audience both by demographics and psychographics. What does it say about you if you drive BMW and read the Telegraph… or if own an allotment and make jam? All this information has been condensed down into a set of number. This approach was proven in the commercial market, CA weaponised Psychographics for electioneering…

Psychographics gained prominence in the 2016 US presidential election since both Hillary Clinton and Donald Trump used them extensively in microtargeting advertisements to narrow constituencies.

So CA’s “This Is Your Digital Life” basically provided a mainline feed into pyschographic data. But CA also collected data on voters using sources such as consumer behaviourinternet activity, and other public and private sources. According to The Guardian, CA used psychological data derived from millions of Facebook users, largely without users’ permission or knowledge. Another source of information was the “Cruz Crew” mobile app that “gamified” election campaigning by giving points for the number of political social media messages circulated by the player. But more than that it tracked physical movements and contacts on the player’s smart phone and so invaded personal data more than any previous electioneering method.

Alexander Nix, chief executive of Cambridge Analytica, October 2016, said “Today in the United States we have somewhere close to four or five thousand data points on every individual … So we model the personality of every adult across the United States, some 230 million people.”

CA’s data analysis methods were to a large degree based on the academic work of Michal Kosinski. In 2008, Kosinski had joined the Psychometrics Centre of Cambridge University where he then developed with his colleagues a profiling system using general online data, Facebook-likes, and smartphone data. He showed that with a limited number of “likes”, people can be analysed better than friends or relatives can do and that individual psychological targeting is a powerful tool to influence people.

This aspect of facebook-likes is absolutely key and – as far as I can tell – is missed in most write-ups of the Presidential Election 2016 and GE2019 Fraud.

Facebook “Likes”

Most, but not all. It was discussed extensively in 2018 by CBS which states “Facebook ‘likes’ can signal a lot about a person. Maybe even enough to fuel a voter-manipulation effort like the one a Trump-affiliated data-mining firm stands accused of — and which Facebook may have enabled. The social network is under fire after The New York Times and The Guardian newspaper reported that former Trump campaign consultant Cambridge Analytica used data, including user likes, inappropriately obtained from roughly 50 million Facebook users to try to influence elections.

The issue of the addictive nature of facebook and the dopamine hit when someone “likes” your post is well known. So how important is that “like” if it just came from a bot? Can they even do that? Yes.

Technology to Support CA

If the data collected by CA was all performed by party workers then would it all be bad? Probably not: doorstepping in elections trys to collect similar type of data. But CA introduced the mechanism to do this quite automatically, without permission, by impersonation and by the the “backdoor”. Besides the (illegal) aggregation of data from a various sources, this is the type of technology that CA used in order to recognise and give facebook “likes”:

  • Artificial Intelligence (AI) – Sentiment Analysis – this sort of AI can read thousands of posts and determine whether any particular post supports the camapign, against it or whether it is just another cat video.
  • Robot Process Automation (RPA) This allows a series of automatic actions to occur online, for instance: read a facebook message, work out if the sentiment supports the camapign, clicks the “like” button.
  • Bots and Sock Puppets – basically fake accounts – either a Bot which is a fake account which performs a repetative RPA action, for example, “liking” a facebook message; or a sock puppet, a human controlled fake account, that can enter into hundreds of discussions online dissing the opponents and/or talking up the camapign with prepared slogans.

The set up and running of this technology requires a huge amount of capital intensive investment (this is where the rich, organised “few” outgun the poor, disorganised “many”). So democracy is now a hidden war between people-powered electioneering (“the people”) against a limitless army of hidden robots controlled and funded by a few billionaires. This technological army is not even in party headquarters but can be outsourced to friendly front organisations, commercial organisations or even foriegn powers.

The Hub

The only thing needed for any political HQ is the data collection hub. HQ will ensure the right campaign messages are being fed in a way that is compelling… to a “plan”. That requires technology again but it is cheaper and readily available off the shelf in the form of a Customer Relationship Management (CRM) system. This a tool to manage an organisations interaction with current and potential customers – everytime you phone up a major corporation nowadays you are being managed by a CRM system such as Microsoft Dynamics CRM, Salesforce or SugarCRM. Now replace the word “customers” with “voters” and the tool works just as well. Just look at the (proven) results of CRM and see how they apply to election campaigns.

  1. Enhanced ability to target profitable customers. (Replace “profitable customers” with “likely voters”)
  2. Integrated assistance across channels. (Use of Bots, Sock Puppets, Newspapers or leaflets to promote your propaganda)
  3. Enhanced sales force efficiency and effectiveness. (Replace “salesforce” with “campaign staff”)
  4. Improved pricing. (Instead of pricing, think “extract donations”)
  5. Customized products and services. (Tailor message for particular poltical concern: health, environment, business etc)
  6. Improved customer service efficiency and effectiveness. (Improve approval ratings)
  7. Individualized marketing messages (also called plans). (A set of messages just to engage each individual voter, there can be multiple plans depending on the voter’s concern)
  8. Connect customers and all channels on a single platform. (A complete view of Voter Intentions).

Architect of the Vote Leave and Conservative GE2019, Dominic Cummings, called his central database of voters, the Voter Intention Collection System (VICS). It is described in his blog. He describes how he developed ads for social media, trialed them and targetted them. The data feeding VICS was both “conventional and unconventional” – from that we can assume conventional = demographic data – freely available to political parties – and unconventional = psychographic… as described above – illegal. Illegal to the price of $5billion just to facebook alone for allowing a loophole in its software. How much more illegal is it for people to deliberately exploit that loophole?

But wait there’s more…

Just consider the power now available to the rich elites with such technology at their disposal:

  • A list of every voter.
  • With social media data then the on-line accounts can be linked to voters in the electoral register probably in 80% of all cases.
  • This enables the identification of all people strongly aligned with campaign messages and will vote.
  • And the identification of all people strongly opposing the campaign.
  • This identifies the battle ground! The non-aligned people.

Pyschographics help sort out the battleground. Since the social network shows who is friends with whom, then the probability of voter intentions can be calculated with different levels of certainity… until you have complete and utter situation awareness of how people will vote. People do respond to the information they’ve received but if all the information is biased and plays into pre-set grooves enabled by the mainstream media then, people respond collectively in herd like behaviour. Dominic Cummings tested his “messaging” in carefully selected groups. When the message had the right effect, he sent out targeted political ads and using the AI and Big Data analysis re-calculated the expected voter intentions so that he could predict an 80 seat majority for the Tories. And he got an 80 seat majority for the Tories. This level of estimation precision requires computers. And probably coercion (see below). The Conservative Party was able to deliver an astonishing efficiency at delivering seats in 2019: One seat for every 38264 votes (a 10% efficiency improvement over 2017) while the LibDems were amazingly less efficient: one each seat for every 300,000 votes, a 50% decrease in efficency. And, unlike Jo Swinson travelling the country in a bus (and even losing her seat), Boris Johnson never really needed to go out and campaign or even do many TV interviews.

Minimum Fraud / Maximum Outcome

There are further tools in the toolchest. Having complete situation awareness allows other useful things:

  • it identifies marginal consistencies.
  • it identifies people that are unlikly to vote
  • it identifies people that will be using a postal vote
  • it identifies people that are misaligned with the electoral registers

We know postal voting fraud exists and is widespread. Complete situation awareness of voter intentions now allows two useful forms of election fraud, which can be set at the minimum level that arouses the least suspicion:

  • Voter Suppression for those people of the wrong demographics and pyschometrics that are misaligned with the electoral register and/or registered for postal votes. (eg postal votes not delivered, arrive late or invalidated). Voter suppression is regarded as a non-crime – the voter is always blamed for any administrative error.
  • Ballot Box Stuffing, by postal votes, impersonating people that are unlikely to vote. A virtually undetectable crime!

What’s all the fuss?

Now do you see what the fuss is all about?

  • No need to campaign
  • Lower campaign costs (as long the computer system costs can be hidden)
  • No need to be held to account for any promises or policies
  • Set up the perfect way to secure a seat with the minimum level of fraud – so small that it is hardly detectable.
  • Confidence of predicting the election result nationally (100% accuracy)
  • Confidence of securing any particular local seat (as long as there is a high level of postal votes!)

Sheeple. Are you ready for this?

Situation Awareness or “How to win an election without really trying!”

How could the Tories be so certain that they would win a General Election? The answer is “Situation Awareness”. You just need to read up on Dominic Cummings’ blog about his Voter Intention Collection System on how it was used in the Vote Leave campaign and how it was re-used for GE2019.

On October 24th, 2019 – there were fears of electoral chaos…

General election: Proposal for snap vote in December sparks fears of electoral roll chaos
Warnings of polling station shortages and administrative mix-ups amid mounting speculation of pre-Christmas poll

“Britain is facing the possibility of electoral chaos if Boris Johnson pushes ahead with plans for a December general election…. A pre-Christmas snap election would force officials to use two different electoral registers to manage voting, potentially leading to confusion and delays…. The electoral roll is updated annually on 1 December, meaning the new register would be in place in time to be used for a possible election… However, polling cards for an early December election would need to be sent out in November, meaning one electoral register would be used for sending the cards and another used for the actual election… It means that people could see their polling cards, which include details of where they should go to vote, sent to their old address… Peter Stanyon, chief executive of the Association of Electoral Administrators, which represents returning officers, said a poll in December would… risked people being wrongly marked as having voted when they had not.

Would this bother the Conservatives? Obviously not, who are most likely to one of the 9 million voters elible to vote but not correctly registerd? The poor (in rented accomodation) or youngsters – both demographics likely to vote Labour. As the article says “You shouldn’t have to opt in to your right to vote. As the Electoral Commission note, we need to move towards automatic registration now, starting with being able to check you are registered online, and being able to register whenever you engage with government bodies or services. We know this works from other countries.” So the information is available of where the voters are and whether they mismatch the electoral rolls… I bet this information had been captured for complete situation awareness of Voter Intentions! It could even by itself explain the drop in the number of Labour votes…

But there is more. Chaos does not increase confidence… only order and control provides confidence. November 2019, Tories predicted an 80 seat majority. A lousy campaign later, #Trustgate, #Wreathgate, Boy on Hospital floor, #Fridgegate etc Dominic Cummings #VICS system gave them exactly an 80 seat majority. That is some “situation awareness”! This level of estimation requires computers especially since the Tories had no ground camapign in the north! They had every voter identified, every voter’s politics analysed by computer, every intention to vote predicted or persuaded or impersonated or blocked. Rigged? You bet!

Election Fraud GE2019

No-one seems to be talking about this yet… GE2019 and the use of Social Media Sock Puppets and Psychological Operations, Postal Vote Fraud and Voter Suppression. This wasn’t so much a political event, trading policy ideas on a level playground, but more like a military operation with one side armed to to teeth with tanks and fighter jets with the other with handbags.

Background

The Tories were in a real mess in October 2019 running a minority government. Johnson was goading Corbyn to call for an election. Johnson obviously had a plan for the election and it had nothing to do with policy or a vision of the future for Britain, it was just three words “Get Brexit Done” – whatever that means (as far as I can tell it is just to let Tory Donors off the hook from tax avoidance scams).

The Tories ran a pretty dreadful campaign. Johnson was Boo-ed on many of his public outings and his team were a shambles, offending Greenfeld victims, NHS workers and patients, and dodging any questions regarding climate change or the environment. Johnson was defending a marginal seat with only 4000 odd majority, but did not even attend hustings in his consituency. A similar story can be found in constituencies in the north of England (soon to be won by them in the GE) where there was precious little evidence of a Tory campaign on the ground.

Corbyn was met with huge crowds of supporters, chanting and cheering wherever he went. But this was not reported in the Mainstream Media, prefering, it seems, to attack him personally, and Labour in general, with “weaponised” lines of questioning regarding anti-seminitism (a problem that affected Labour at a rate of 0.01% of its membership). The BBC – supposedly impartial and in “purdah” – went out of its way to edit video to make Johnson look good, covering up on the Jennifer Arcuri scandal, the withholding of the “Russia Report” and the social media manipulations, particulalry the sock puppet accounts, over the boy-on-the-floor in Leeds Hospital. https://www.yorkshireeveningpost.co.uk/news/read-this/general-election-2019-leeds-general-infirmary-fake-news-flooding-social-media-explained-1337095. These sock puppet accounts (as reported by the BBC) demonstrated the immense social media capability available to the Tories.

Various commentators noted that the gap between Conservatives and Labour was narrowing mainly because the polls had not accounted for the Youth Quake, dying Brexit voters and Campaign disasters (such as Fridge Gate). There was the issue of Labour voters moving to the Brexit Party but both the LibDems, Brexit Party and Tories had a dreadful campaign and many people expected a shift to Labour and at least a hung parlaiment.

General Election 12th December 2019

So on the day of the General Election, there were extraordinary queues at polling stations in many consituencies: Young Voters coming out in force. But at 10.30pm the exit polls were announced and the BBC was 99% accurate to the actual result… which is rather strange since the postal votes were at an extraordinary high – double the number in 2017 and not counted towards the exit poll. Did the BBC have other data to work with other than the exit poll information?

GE2019 results cf 2017

According to the official figures, while the registered voter population had grown from 46m to 47m, turned out had dropped slightly from 68.7% to 67.3%. These figures seem to be a bit of puzzle… considering not only the Youth Quake, Polling Booth Queues but also the reported doubling of the postal vote from around 16% in 2017 to 32% in 2019 – these figures are still being collated since official figures will not be fully available until Jun 2020.

The official explanation is that former Labour Voters moved to Conservative (and Brexit Party) and while this story has credibility according to these figures, more striking is the move from Labour to LibDem, Green and SNP. Perhaps the real story is how efficient the Tories had optimised their vote in the very unfair First-Past-The-Post election system, allowing them to pick up one seat for every 38264 votes while the LibDems were amazingly less efficient moving from 200,000 in 2017 for each seat to 300,000 in 2019. It seems that the Conservatives with only 100,000 or so paid up party members had out-campaigned Labour, the largest political party in Europe with 500,000 party members, through the use of capital-intensive technology.

VICS

Before the EU Referendum, Dominic Cummings had created a system for Vote Leave called the Voter Intentions Collection System (VICS). This is well documented in Dominic Cummings own blog. https://dominiccummings.com/on-the-eu-referendum/ It provides a statistical probability of each and every vote and voter, fed by data sources “conventional and unconventional”.

For Vote Leave this Big Data system was fed from information from facebook: information on 2 billion (yes, 2 billion) people collected by Cambridge Analytica. There is an ongoing case about the fraudulent collection of this data and still further speculation whether this data was available to Dominic Cummings in 2019.

In 2019, Dominic Cummings claimed that voters were persuaded to change their attitudes to people, party and desire to vote by targetted adverts. Facebook allows targetted ads including by location, demographics (age, gender, education, job title), interests, behaviour and connections, include people who are connected to your people or events, or to exclude them. So the Tories had the ability to target people that were likely to be swayed by their ads. Cummings claim they tested the ads on small groups, fine tuned them and then pushed out to larger groups. The ads were largely attack ads against Corbyn or “Get Brexit Done” messages with overtones of racism. The ads were targetted towards football supporters and football supporter mentality… which is largely pretty tribal if not racist…

Sock Puppets

However to really guage voter intention, VICS needs more than just targetting information it needs, feedback and engagement. Social media, particularly facebook, is the tool of choice. Sock Puppet accounts and bots were enabled to do a number of things:

a) identify a person who was likely to vote for a particular party – if confirmed views, would they be voting?

b) identify people who could possibly be infulenced “floating voters”, and if infulenced would they vote?

c) identify people who were unlikely to vote at all

In particular, besides identifying voter’s political view, the sock puppets and bots could re-inforce tribal behaviour and commitment to vote by either “liking” posts, engaging in chit-chat and obtaining such confirmation. As mentioned above, the Tories had demonstrated a massive capability for its sock puppets and bots but it is unlikely that these were directly controlled by Tory Party operatives but either UK Intelligence Services or Foriegn Intelligence Services – the most likely being, not Russia, but USA and Israel who both have massive capability for such “Pyschological Operations“.

IDOX

IDOX systems are used to used to manage Postal Votes in around 80% of UK constituencies. It does not count votes but managing the requests, sending out voter packs, verifying the packs and managing the electoral rolls. This is gold dust info into a system like VICS. So was the data passed deliberately to Tory CCHQ? Maybe not necessary – just hacked! Rabid Tory Peter Lilly was a former director of IDOX, he could have passed on details to people that could then easily hack the IDOX systems. We know for instance that each IDOX installation was separated (so no central database) and I’m sure, “good security controls” but each system was also capable of being connected by wi-fi and/or the internet and so, of course, wide open to any determined organisation to hack into it. (See picture). The Institue of Statecraft for example, would easily be able to achieve this.

Postal Vote Fraud

Various efforts are now in play to investigate the postal vote fraud but this is nothing new. There was a postal vote court case in England in September 2019 and suspected postal vote fraud in the Scottish Referendum For GE2019 early information indicates that both Safe Tory Seats and Marginals won by the Conservative, had high Postal Ballot percentages, double the usual figures.

The VICS data would have indentified those people unlikely to vote. These would be the perfect people to impersonate for postal vote ballot stuffing.

Voter Suppression

Finally there is evidence of voter suppression, particularly of youngsters, people living abroad and new British Citizens, (and there’s more on the way). The evidence for this is available all over Twitter. Another group, confirmed non-Tory voters, could have been identified by VICS data and these voters could have been somehow twarted from voting (by a variety of means).

Situational Awareness

The different methods to achieve the goal, a conservative landslide, would have been diverse in order to stop any one particular method being fully understood and the fraud exposed. All that really was required was “Situational Awareness” – a military term that is now widely used within business – to provide a complete overall picture of progress towards the goal. In the conversative case, this was provided by Dominic Cummings’ Voter Intention Database. Is that system illegal? No… Well, actually it depends on who paid for it and how it was accounted for in campaign finances… and we won’t know that until June 2020. Oh, and if using Social Media Data it probably broke GDPR rules governed by the Information Commissioner Office. (The ICO seems to be very quiet on this matter).

Follow Up

More information will be written up as it becomes available. This blog will be used to store information and links provided for use on social media…

Electoral Law is lagging behind in regulating Big Data and other technological advances that destroy real democracy…

2017 – 13.63m vote Tory…..Hung Parliament

2019 – 13.96m vote Tory…..Landslide victory

Something is not right.