How To Lie With Numbers
It takes less effort to lie without numbers, but there are now more numbers and more ways to lie with them than ever before. Poor Reverend Bayes, who understood the true meaning of "evidence".
This is a Golden Era for the huckster. Here are some of the many tricks you can use if you're one of them. Few are very difficult, and all you really need is no ethics.
- Your pitch will be more convincing if it has numbers. If you don't have any numbers, just make some up. Rankings, for example, impress many people and are easy as pie to fabricate. Just arbitrarily select some criteria and subjectively weight each criterion. Claim the weights were determined by AI. Then get your friends to score the companies or whatever it is you’re ranking on the same criteria. Sum the weighted scores and, Bingo, you have expert judgements! By adjusting the weights, you can get any ranking you want.
- People respond inconsistently to logic and evidence. Use emotional appeals, sprinkled with numbers that appear to support your case.
- When making the case that there is a need for whatever it is you're selling, generalize from the exception. Make rare events seem more commonplace than they really are. Good way to drum up VC money.
- Deliberately confuse potential with reality. Sure, your gizmo might one day serve a real need, but it has no track record yet. Use projected trends from official sources as smoke screens to make it seem like it is serving a real need NOW. It doesn't matter if these trends are relevant to what you are selling. "By 2025, there will be XXXX billion connected devices!" works just about every time for just about anything.
- Build your sales pitch around the few cases when whatever it is you're pitching did work. Showcase these, and make sure not to mention the dozens of times it flopped.
- Construct your argument so that anyone who questions it is made to seem ignorant, stupid or naïve. Numbers can really come in handy here, especially when they are bogus.
- Establish your lies as the truth through constant repetition.
- You can be confident that few people will scrutinize your claims. Only a handful will remember all the predictions you made in the past that turned out to be wrong. If they do, revise your predictions so that they appear to have been accurate.
- Recall that few journalists have had much training in statistics or have deep knowledge of the subjects they report on. They also have convenient memories about what they have reported in the past.
- Understand that people will believe you if they want to believe you. The trick is to make them want to believe you.
- When your method or technology has been shown not to work, cite "new research" as vindicating it: "You shouldn't have believed me then, but you can believe me now. Trust me." in different words.
- When your method or technology has been shown not to work, make the false claim that "No one does that anymore.That's the old way." to discredit your critics and wiggle out of the jam.
- Complain about "hype about hype."
- Conversely, admit there's been a lot of hype about what you're selling. Sound exasperated. Then rephrase the hype.
- Cite discredited research.
- Cite discredited research as discrediting the research that discredits your research.
- Cite experts. Lots of them. It doesn’t matter if their views actually support your position. Throw in some famous authorities and some people no one has heard of. This will make you look like an authority yourself, intimidate your critics and make them waste their time looking up your citations. Avoid citing books that are freely downloadable and papers that aren’t paywalled.
- Cite the handful of experts who happen to agree with you and claim their views are what nearly all experts on the subject believe.
- Misrepresent the views of experts or quote them out of context, as needed.
- Distort your critics' case. Straw men dot the landscape these days, but there's always room for one more.
- Quote academics who have a commercial interest in your claims. Accuse your critics of having an “agenda” or financial conflicts of interest to conceal your own agenda and financial conflicts of interest.
- Deliberately confuse theory with evidence.
- Use lots of jargon. Use words like "Bayesian" and "Meta-Analysis" that are now fashionable but misunderstood by most people "Disruptive" may have seen its day but can still come in handy. No need to define what it means.
- When you've been exposed, respond with long, complicated rebuttals that appear authoritative but are hard for laypersons to see through. You don't have to say much of substance. Make your opponents seem like dummies who just don't get it.
- Simple arguments are often effective. However, many people are impressed by things they don't understand. Depending on the audience and what you're pushing, either oversimplifying or mystifying may be the better tactic.
- Cherry pick research that supports your argument. This is most effective when no one has attempted to reproduce or replicate this research.
- Cherry pick data. Include or exclude variables as needed.
- "Adjust" data so that it supports your claim. Use percentages creatively, for example saying that your sophisticated targeting increased conversion by 50% when the actual increase was from 0.001 to 0.0015.
- To "prove" that your gizmo can accurately predict sales, for example, retrospectively mine data and only mention the hits, never the misses. Make it seem like you made these "predictions" before the fact, not after. Who will know?
- Simulate data and pass it off as actual data.
- Use simulations programmed with the assumptions you're trying to prove as proof these assumptions are true.
- Cherry pick models. Model the data until it confesses!
- Deliberately reverse cause and effect.
- Use new and unproven methods but sell them as cutting edge. Or, use old methods with a new name you've made up. Better yet, call anything AI!
- Use data visualization to exaggerate small differences or to conceal things you need to hide.
- Only cite part of a data series - ideally just two points in time - that are consistent with your claim.
- Quote numbers and research out of context. Quote selectively from research.
- Humans have a strong tendency to think in black and white terms. Use this to your advantage. For example, make it seem like anyone who questions your “innovation” is a dinosaur.
- When backed into a corner and in risk of being exposed, use misdirection. Fire back with data or arguments that are entirely irrelevant in order to confuse and distract your opponents.
- Humans often see patterns in randomness. Use this to your advantage.
- Many people mistakenly believe random means even. Use this to your advantage.
- Make creative use of benchmarks. For example, present bad performance as if it were standard performance. Concentrate on failure and don't mention the majority of cases when current practice or technology works just fine.
- When you can't rebut an opponent's argument, resort to ad hominem tactics.
- Set the bar impossibly high for your opponents and conveniently low your yourself.
- Use statistical significance to your advantage, for example when effect sizes are small but sample sizes are large.
- People want to believe in miracles, so claim you can do what no one else can do. Conceal the details of what you actually do and that it doesn't work.
- Distort criticisms of statistics or use them out of place. For example, don't let on that the normal distribution is but one of dozens used by statisticians.
- If your critics make a trivial error - even an innocent typo - use that to discredit their entire case.
- Play Whack-A-Mole with your critics. When they hammer a point you’ve made, respond swiftly with another, unrelated remark. When they hammer that one, quickly respond with a different argument that is unconnected to the first two. Keep popping up randomly all over the place. This is one way to confuse and wear down your opponents, especially those who think logically and play fair.
- If someone completely demolishes your position with logic and evidence, perhaps your best option is to forget everything I've said and simply ignore them.
There are many tricks that can be used to deceive us. Please be on the lookout!!
Original. Reposted with permission.
- Vital Statistics You Never Learned… Because They’re Never Taught
- Statistical Modeling: A Primer
- Time Series Analysis: A Primer