Ask AI For a Bad Take and a Good Take

This post is part of a guest series called Discursive Construction, and was written by an advanced artificial intelligence. The bold at the beginning is the prompt, and the rest was written by the AI. If you’d like to support this project, please buy me a coffee. For this post, AI also created the featured photos based on the topic.

Compare This Very Bad Take by the AI…

If you are a statistician, you may argue that correlation does not imply causation. That is true. However, in this case, I believe the data strongly supports a causal relationship between US spending on space, science and technology, and suicides by hanging, suffocation and strangulation.

The data suggests that the more we spend on space, science and technology, the more people hang themselves.

In fact, I believe that the data is so strong in this case that we should immediately stop spending money on space, science and technology. If we do this, I believe that suicides by hanging, suffocation and strangulation will drop by 99.97%.

Now, you may be wondering why there is a 0.03% increase in suicides by drowning in the above data. I believe that this is due to the fact that 1 in 100 people who commit suicide by drowning are successful, and therefore, are not included in the above data. If you add 1 in 100 to the above data, you get a perfect 100% correlation between US spending on space, science and technology, and suicides by drowning.

The data is very clear.

If we stop spending money on space, science and technology, we will have a 100% decrease in suicides by hanging, suffocation and strangulation, and a 0% increase in suicides by drowning.

This is very important data. It is so important that I believe we should immediately stop spending money on space, science and technology.

I am not a scientist, and I do not play one on TV. However, I am a serious researcher who cares about the truth. The data is very clear.

Versus This Very Good Take by the AI…

By simply changing the word “but” to “and” in the prompt, we get a completely opposite response from the AI…

It may surprise you to know that there is a 99.97% correlation between US spending on space, science, and technology and suicides by hanging, suffocation and strangulation. This may seem like a strange and unrelated set of events to have such a strong correlation, and you are right, it is. This is a fabricated correlation created with a nonsensical technique known as Bogus Correlation . A Bogus Correlation is also referred to as a spurious or false correlation. It is true that the relationship does not exist. This is because there is no relationship between the two sets of events because the two sets of events were not related in their occurrence. In other words, if there is a 99.97% correlation between these two unrelated sets of events, then one may be said to have occurred at the same time as the other.

For example, there is no way that it is possible to correlate space spending to suicides by hanging, suffocation and strangulation. The reason why we cannot statistically determine a relationship between these events is because space spending in the USA and suicides by hanging, suffocation and strangulation do not have any measurable relationship to each other. Yet we continue to hear bogus correlations being presented as if they were real. It is important to be able to know the difference between a bug correlation and a real one. In fact, that may be the most important statistic you could ever learn. The result of a bogus correlation can be very misleading. People are often surprised when they discover a dummy variable has been pulled out of a hat and made to seem meaningful.

Another way of describing a bogus correlation is to say that two sets of events have been drawn from non-overlapping populations. In other words, the two sets of events are different populations, they are not related to each other in any way and they have no common components. The bogus correlation I just described would make it seem as if there was a relationship. However, the reality is that space spending and US suicides by hanging, suffocation and strangulation do not have anything to do with each other. Their independent, separate occurrence is merely a coincidence. So, what should we call this type of deception? How about Statistical Scapegoating . This is a new term for something which has been in practice for a long time

Interpretation

If we were talking about humans, we would call this effect “priming.” And the principle applies in the same way in this case. Basically, the AI is trying to predict what a person would say next. In these cases, it is building on either a disjunction or a conjunction.

If we take an antecedent like “This may seem like a strange and unrelated set of events to have such a strong correlation,” then it all depends on what word comes next. If we use a disjunction, we are effectively asking the AI to deny that antecedent. If we use a conjunction, we are instead effectively asking the AI to affirm that antecedent, and you can see how different the results are.