2. Introduction
• In this lecture we will consider a very contemporary
issue in social media – fake news and its impact upon
the Brexit vote and the US Presidential election.
• In order to consider fake news how however we need
to look at some of the more fundamental aspects of
social media:
– big data;
– the use of inferential data analytics used on social media;
• We have saved these until this lecture as they draw upon some of
the ideas introduced earlier in the course
– How fake news is used.
3. Data
• The leveraging big data requires the use of enormous amounts of
data about individuals, their choices and their networks of
interaction with others.
• While much data was collected in the pre-digital era (we have
collected census data since biblical times and earlier in parts of East
Asia) it was often scattered, disjointed and not integrated.
• Individuals and organisations would collect data but it was never
integrated into larger sets or used in a strategic manner.
• Recording the information in a digital manner makes it possible to
combine the information, search and ‘mine’ it in ways not possible
before.
• One of the aspects of digitality.
4. Sources of data
• Lots of ways in which data about us is collected.
• Through engaging with digital media we give up information about ourselves
either willingly or unwittingly.
• Willingly:
– We give it away for a better, more efficient or accurate service, a better service.
– We trade it of something - free access to wifi, online content 2016 OFCOM
survey 2 in 5 will trade personal details for content, most likely in 16-24 yr olds.
• Unwittingly:
– do not realise the terms and conditions mean or details are passed on – 2016
OFCOM survey only 1 in 5 admit to reading online Ts&Cs in detail.
We you use free wifi you will often find a
statement in the Ts and Cs that they will
register your mac address. This is the device
specific address. This can tie the device, and
therefore your browsing and exact location to
you. This will then be linked to loyalty card.
5. Searches, likes and shares
• Action on social media produces data.
• The terms we use to search, the videos we watch,
the things we ‘like’ and share all, even the phone
calls we make with integrated devices (if you
make a call on your mobile to a company they
may well appear in your newsfeed) all contribute
to our personal (and friends and organisations)
data footprints.
• Once phone mac address is obtained movement
can be recoded as well.
– Happens in shops and shopping malls.
6. Social media personality quizzes
• Lots of people have filled in
personality quizzes on social media
which links your personal information
(name, age, sex and other
information) to the output.
• This data is channelled into large
databases and then ‘mined’ and used
to build OCEAN personality type
models for people –
• Ocean is a five-factor model is
comprised of
five personality dimensions (OCEAN):
– Openness to Experience,
– Conscientiousness,
– Extraversion,
– Agreeableness,
– Neuroticism
7. Data for sale
• There is lots of consumer data available for sale.
• Companies trade information that establishes you
as a commodity.
– Credit scores (Experien) and consumer data can be
bought and integrated with other systems to create in
depth profiles.
– Electricity, mobile, data use, club cards (shopping
details), credit card purchases.
• This information is packaged up and sold on.
8. Worth
• On its own such data is not worth very much.
• Trade in individual data points about a person
costs only pennies.
• Also a growing belief that people are getting fed
up with this and will resist giving it away for free.
• OFCOM survey reports decline in willingness to
give away information and apps are appearing
that will reward you for giving away information
or taking personality quizzes.
9. Two ways in which data becomes
valuable
• Single data points are
of little use they need
to be combined with
others to.
• Financial leverage of
data comes in 2 ways:
– Recombination
– Big data
10. Recombination
• Here organisation pull together data from
different sources and use specialised software
to construct their own ‘whole personal data’
sets and even more.
• An indication of the value of individual ‘whole personal data’
sets (fullz) and other personal data is that they are available
through the dark web at a cost – varies in cost from about £25
to £500 (sold in Bitcoins) per individual. Use for identity theft.
• May start with a customer database (Tescos
club card has 15 million members, Nectar 16.8
million) or post code database to which they
add layers and data about individuals.
– Combining data bases through data mining activities.
– Filling in blank data points and adding new layers.
– Social media adds lots of soft data.
12. Big data
• Individual data gets more valuable when combined with
lots of other data points both within a case and across
cases.
• Once integrated it becomes possible to detect patterns and
draw out insights which can be acted upon.
• Using lots of information we are able to predict a result.
• We make inference based upon other examples.
• We are not saying there is a cause and effect relationship.
• Instead a positive correlation between two or more data
points that allows us to determine the probability of a
third.
13. The more cases and data points we get the higher the
probability / more accurate the ‘other people also
liked’.
Dave Jane Steve Agnes Peter
Age 22 22 22 22 22
Favourite
romantic
film
Titanic Titanic Titanic Titanic Titanic
Favourite
TV cartoon
Simpsons Simpsons Simpsons Simpsons Simpsons
Favourite
Harry
Potter book
Goblet of
Fire
Goblet of
Fire
High
probability
will be
Goblet of
Fire
Goblet of
Fire
Goblet of
Fire
Amazon database.
14. What is done with the data?
• The majority of this mining occurs for advertising
reasons in pursuit of ‘direct marketing’.
– Marketing involves getting your message in font of
people who may be persuaded to by your product.
– However the foot print of a specific media may not
overlap with your target audience and as you pay for
all viewers any viewer of your advert who is not in
your target group is a wasted view. It would be better
if only those in your target group saw your advert.
• Direct marketing is where you only hit your near
exact target customer.
15. Data mining combined with social
media
• Data mining allows companies to identify very specific populations
they want to hit with advertising and to hit them in new platforms.
– Basic Facebook marketing allows lots of specification by location, age,
interest etc
– Custom lists can be created from mailing lists that are then imported
into Facebook.
– ‘Dark posts’ adds developed to appear in people’s news feeds, do not
appear on originating pages news feed, only in the recipients feed
(developed to reach new people but not duplicate information for
those who already know it).
– Because these adds are targeted they can be tested for effectiveness
far better than any print add ever could (thousands of variants are
used to determine which is the most effective).
• None of this very new and has been going on for 5+ years.
16. New stuff
• What is new(ish) is the integration of the
customer database plus OCEAN data with
Facebook data.
• Producing very detailed information about
individuals and very strong predictive power
of aspects of behaviour.
17. The power of combing ‘likes’ with
other data…
• Research indicates that the patterns of Facebook ‘Likes’ can
very accurately predict:
Michal Kosinskia, David Stillwell, and Thore Graepel (2013) ‘Private traits
and attributes are predictable from digital records of human behavior‘,
18. Of use to?
• Such information is very useful to marketers.
• The can tailor advertising, predict behaviour
and to a degree and even shape it.
– Life event, article about appropriate course of
action, advert.
– Also of interest to politicians…
19. Cambridge Analytica and Trump
• Big data used to influence elections.
• Trump criticised and laughed at for not
spending much on press campaign.
• He spent it all on Facebook and Twitter.
– He spent half of what HC spent on TV adds.
– Three times what she spent on digital.
• And he employed Cambridge Analytica…
• https://www.youtube.com/watch?v=n8Dd5aV
XLCc
20. What they do
• Simply put Cambridge Analytica build
probability models of how people vote.
• They then look at what they can do to
influence that.
• One of the main ways they do this is through
using fake news.
21. Fake News
• Sites and Facebook pages which present things which are not true as news.
• Look like real news sites eg ABC News a reliable news site – this is
http://abcnews.com.co/obama-signs-executive-order-banning-national-anthem/.
• Many were produced in Macedonia. https://www.channel4.com/news/fake-
news-in-macedonia-who-is-writing-the-stories
• Had lots of scare stories, mainly right wing ones.
• Written to be shared, would derive advertising revenue – some made an awful
lot of money during busy times.
22. News commentators
• Commentators shared fake news stories that
fitted their political agenda which significantly
extended their reach (and increased ad
revenue for the creator).
• Fake news sites were also inserted into certain
people’s news feeds.
• The people were selected according to their
data profiles to sway their voting intentions.
23. In the UK…
• The UK is not so subject to fake news as the news from
genuine sources tends to be quite fake anyway.
• The most shared story during the EU referendum was this
one from the Daily Express.
• Later found to be completely untrue.
• But this was still shared and inserted in numerous selected
newsfeeds.
24. Conclusion
• Lots of data is floating about and we often
give it away.
• It is used to create profiles about us and to
build big data models.
• The profiling it affords is very valuable to
marketing people.
• It is also used to influence and drive politics.
• Fake news is used as ammunition in the gun of
inferential data…