Misinformation spreading across social media channels has become a widespread problem. A growing trend is to post fake news, “alternative facts”, or other misleading opinions, either to promote a political agenda or simply get more clicks, engagement and sales.
The problem with lies are they often cause people to make rash decisions based on the wrong information.
Don’t believe everything you read
The latest industry news on analytics would have you believe that Cambridge Analytica (CA) used psychographic marketing models to win the election for Donald Trump. They claim to have surveyed 1 million US citizens and expanded the findings from this profiling to every voter in the USA. Allegedly the company behind this have 4-5000 data points on every single adult.
There are a number of flags when I see claims like this based on over 2 decades of working with statistical information and measurement.
- How can you reliably have 4-5000 data points on every single adult in the USA if you’ve only got 1 million responses to your survey? Just because someone’s profile like age, sex or demographics can be correlated to another person on Facebook doesn’t mean they have the same profile. You may have some segmented similarities but you certainly don’t have 4-5000 data points on every adult.Its true you can take psychological or demographic information and correlate it with political direction. But this data is available to everyone else (which is why everyone else predicted about a 20-25% chance of Trump winning). What CA are saying is that that they’re somehow accessing personality traits through Facebook data for the other 100s of millions of Americans they didn’t assess via a questionnaire. At best that’s extremely unlikely to work. What people post on Facebook isn’t always what they would do in real life and sentiment analysis at scale is extremely unreliable. Then there is the challenge of bringing the data together. How accurate can you be when you’re trying to match a Facebook account with a name on the electoral role?
- CA represented Ted Cruz before working for the Trump campaign. At that point Trump’s data science operation was a $1500 website. So in reality the data science from CA was wiped out in a few short weeks by a builder with a $1500 website.
- CA also gave Trump a 3/10 chance of winning the election. That’s worse than the toss of a coin. And when Analytics companies do a statistical analysis to do predictions you don’t predict anything unless you’re 95% sure of the outcome (or at least that’s the best practice). So its fair to say they were 95% sure he had a 70% chance of losing.
When you read “We have 4-5K data points on every adult in the USA and that’s why we helped Trump win the election” it sounds plausible. But without the context I’ve just outlined you can’t make an informed decision. Hopefully you now understand my skepticism.
This is the problem with fake news. We read a snippet of information that is either a downright lie, promotional, or something which is out of context.
All that happened was that CA bought a lot of targeting information and did a lot of testing of ads across a variety of channels. That is completely normal in the industry and may have had some impact. It may have turned some voters, but it wouldn’t have won Trump the election alone.
So what won Trump the election?
His message. That’s the reality. The US need to learn to understand the motivations on both sides of the electoral divide and come to acceptable compromises.
When Donald Trump talked to coal miners and said…
“We’re going to get those miners back to work. I’ll tell you what. We’re going to get those miners back to work … we’re not going to be Hillary Clinton, and I watched her three or four weeks ago when she was talking about the miners as if they were just numbers and she was talking about she wants the mines closed, and she will never let them work again.”
It really hit the mark. They may have been repulsed (or not) with his other policies, but Trump was making a simple promise to put them back to work. He spoke in a language they understood.
Clinton on the other hand was saying…
“We’ve got to move away from coal, but I don’t want to move away from the people who did the best they could to produce the energy we relied on … I am passionate about this, which is why I have put forward specific plans about how we incentivise more jobs, more investment in poor communities, and put people to work.”
For a miner “we’ve got to move away from coal” means “I’m losing my job.” And what does she mean by incentivise new jobs? Her communication has nuances whereas “were going to get those miners back to work” is a simple, understandable promise.
There were many such communication battles and by being direct Trump won a lot of them.
So how do you spot the truth?
There are a few things you can look out for when you’re reading or watching the news.
- Is there independent confirmation of the facts? Or is there a single point of view? The first reason I don’t believe that CA won Donald Trump the election is because there is no independent confirmation. Even Trump hasn’t said anything.
- Are there attributed quotes? Newspapers can be sued if there are quotes that are out of context or incorrect so if you see quotes from individuals they are likely to be correct, but be careful on context. A quick Google Search with the same quote will show you all the times that a quote has been made. So if a newspaper is playing dirty Google knows.
- Authoritative voices. If well respected industry peers come out on one side or the other about the item being reported there is a chance to encourage debate and find out the truth with multiple perspectives taken into account. Again in the case of CA I don’t see any commentary from any of the industry rock stars. There is one view (theirs) and everyone else is “jealous” of their success. It is like a single scientist saying he’s created a fountain of youth. Until he’s had a number of scientists conclude the same thing then you cannot call it scientifically true.
- Can the argument be made up? If there are a lot of clever sounding words and an air of mystery it might not be true. If it’s not possible to test the idea or verify it 100% how can it be believed? This was another reason Im skeptical of CA. At best it’s promotional work that can’t be proven. At worst its a lie.
- Consider the source. Is the source of the news reputable? Are other reputable news outlets carrying similar reports? I once saw a headline about a World War 2 bomber being found on the moon. Doesn’t mean its true. Again the sources I read all quote CA themselves and come from their own PR. Great advertising for them no doubt but true?
Platforms and Government are catching up
Facebook took a lot of stick after the Trump election and have responded by working with five independent fact checking companies, ABC news, Snopes, Plotifact, FactCheck, and the associated press. If two of the 5 report that the article in question is dubious then they will flag the news as “disputed”. This will be interesting to see on Facebook.
Governments thankfully are also looking to stop fake news from spreading. In November 2016 officials from 11 countries gathered in Helsinki Finland to plan the formation of a centre to combat disinformation cyber-warfare including spread of fake news on social media.
Misinformation is everywhere
It is happening everywhere in every industry and every political battleground for the hearts and minds of the voter.
Arm yourself with simple tactics mentioned and hopefully by the time the next election in the USA gets around people will be wiser about how to spot what’s real from what’s fake.