So this article came up on the BBC News:
It seems that fake reviews are a massive issue for Amazon
This bit caught my eye:
‘Amazon said it was using automated technology to weed out false reviews’.
Using AI to try to combat fake reviews is nifty but can present a whole new set of issues. Fake news is already being fought using Artificial Intelligence, using software like the Fandango Project. So how does this work in the real world?
AI v Fake Reviews
The method works like this: deploy natural language processing to detect unusual patterns of text, writing style, and formatting. Score everything and build a points logic that can throw up red flags, for further investigation by automated decisioning software, linked to machine learning using historical data using a deep neural network, and bosh*.
*possibly. I made this up whilst eating a bagel.
Here are some human (common sense) solutions to the problem:
Require proof of purchase/ownership/use
First off, simply turn off reviews for anyone who has not *actually purchased that item/experience*. This seems pretty common sense. Then wait 30 days before they are able to review it, and make sure they have not been refunded the money, without sending the item back (a common way to pay for positive reviews)
With holidays/experiences/hotels the issue is a bit more difficult. Trip Advisor will let an American human person post a review of a 5 star all inclusive holiday in the Maldives, despite never owning a passport, or leaving the comfort of the home. This is clearly bogus.
Amazon can solve this issue for purchases, as they have the purchase history, duh.
These can be determined, if a person meets these three scientific criteria:
- A regular contributor, with good standing and zero red flags (not *actual* red flags, am not flagrophobic)
2. 2 factor Verified
3. Writes like an actual human person does. (If a review has a frequent use of ‘I’ and ‘me’ it is more likely to be bogus, or the work of a narcissist but likely Sandra from Epping is really Sandip from Bangalore, or a Bot.
Star ratings — a made up nonsense metric
Star ratings don’t mean a thing as one persons 3 star might be your 2 star (ooh get you!) and who would ever give a five star rating for anything? Even the greatest products and services are not perfect. Everything can improve.
There is a seven star hotel in Dubai, and an *actual* 7 star hotel in Milan which has now downgraded your 5 star hotel to a mid range 3.5 star. Also what happened to 6 star? Its almost like they knew that if they made a 6 star hotel, then it would not be long until a 7 star hotel rendered the 6 star hotel superfluous. If your rating goes up to 10, then surely a rating that goes up to 11 is ‘better’.
Netflix got rid of stars and used a thumbs up/down mechanic which works fine. (this worked well in ancient Rome, where gladiators would signify the end of a fight with either a thumbs up or a thumbs down.)
Quick Jimmy Fallon break:
Human reviews are rubbish too
Humans are a bit stupid. Bear with me as I elaborate on this.
Evidence A: The Great Wall of China
No further questions, your honour.
The other glaring issue is that human reviews are dangerously unreliable, impossible to quantify and based around human psychology, which has never really been that kind to humans.
People who have an extremely positive or negative interactions with a brand experience ‘incidental arousal,’ which gives them a strong desire to share their opinions. These strong opinion are loaded with bias, super subjective and achingly personal (narcissistic).
Samples sizes are never really large enough. As any statistics teacher will tell you, all things being equal, the average user rating should be more informative as sample size increases relative to variability.
People lie. Big time. Also that holiday you paid £5k for, are you really going to admit that it was not worth the money, or that holiday you paid £350 for, how do you know whether it was value for money, as that is, at the base level of purchase, the only real metric you can use when comparing anything.
Also personal taste is a minefield. Condider the popularity of the movie ‘Mama Mia’, a ghastly post apocalyptic culture dumpster that fused ABBA with Pierce Brosnan. It was the highest grossing movie in the UK, surpassing Titanic (the movie, not the boat). If Mama Mia was a boat it would have sunk the iceberg.
Humans are in general, a bit daft (see Brexit) and prone to liking a bunch of stuff that is patently a bit daft (see onesies, Crocs and selfie sticks).
Is an honest review from an idiot any better than a fake review? For any purchase that you can subsequently return for a refund? not really. Just buy it and if you don’t like it, then return it. The following is the law on online purchase and mail order(distance selling) in the UK:
Online, mail and telephone order customers have the right to cancel their order for a limited time even if the goods aren’t faulty. Sales of this kind are known as ‘distance selling’.
You must offer a refund to customers if they’ve told you within 14 days of receiving their goods that they want to cancel. They have another 14 days to return the goods once they’ve told you.
You must refund the customer within 14 days of receiving the goods back. They don’t have to provide a reason.
How to buy headphones
Headphones are a great example of why reviews are messed up.
According to a Which study on fake reviews for headphones, all the products on the first page of results sorted by average customer review were from little-known brands and 87% of more than 12,000 reviews for these products were by unverified purchasers.
Seventy-one per cent of the headphones had perfect five-star ratings, while some included reviews for unrelated products such as soap dispensers. One set of headphones made by the brand Celebrat had 439 reviews. All were five-star, all unverified, and all arrived on the same day.
Buying headphones for dummies:
- Ask your friends, colleagues or family which ones they have, why, and if they like them
- Actually read a few online reviews rather than just looking at the ratings
- Don’t ever buy ones with a 5 star review
- Try some on in a shop
- Don’t buy Apple AirPods. “They are future fossils of capitalism” (Vice)
Beware the 5 star curse
Product purchases are most influenced by reviews with an average star rating between 4.2 and 4.5. Products with five-star ratings less so, as there is a level of skepticism. If you buy a 5 star rated product, you deserve to be disappointed.
Ok so time to wrap this up
Online reviews need to get better, but so do humans. Before we bought everything online, we got on fine with analogue shopping
- Buy clothes that suit you, not a Kardashian.
- Buy shoes that last, and that dont hurt your feet, anything else is form over function and nobody likes a hobbler.
- Headphones, dont buy Beats, Apple or a company with no audiophile history and don’t buy a brand that you have never heard of. There is a reason why people buy and respect brands such as Sennheiser, Sony and Shure and that is because they all start with the letter ‘S’, and that is the best letter ever. Ok I made that up. Also buy ones that fit your ears.
- Stop comparing your tastes to other peoples. “One man’s trash is another man’s treasure”. Are you really going to benefit from a review from someone who thought James Bond singing an ABBA musical was worth paying to go to the cinema?
- There is a seat for every bottom. However if you think that the Great Wall of China ain’t so great, you deserve to stay standing.