RownhamHill said:
No offence taken at your disagreement, but I don't agree with your analysis. Take those judgements:
probably guilty (not definitely) - obviously this was a statistical probability analysis, we agree, but to clarify my use of probably here: I definitely did think he was (probably) guilty (obviously notwithstanding I'm not omniscient!). If I'd based my judgement simply on probability of guilt I would have convicted, no doubt.
I think you’re saying you wouldn’t convict just on “probably”. How is that not statistical reasoning? Surely you don’t think statistical reasoning demands one always go with the majority? Surely you can understand that we can distinguish degrees of probability—again, statistically--and don’t assign guilt until the degree is very high? If we were unable to do that, we couldn’t distinguish preponderance of evidence, in civil cases, and beyond reasonable doubt, in a murder trial.
complacent job - I'm not sure how a value judgement about someone's competence is a statistical analysis is it? I didn't estimate whether he'd done a bad job, but made an observational deduction - person x accuses person y of lying; person x presents absolutely no supporting evidence that person y is lying other than he 'probably' was: therefore person x (whose specific job as prosecutor is to present evidence) is doing a bad job. As I say I 'definitely' think that person y probably was lying, but there was a complete absence of evidence presented to back up my own internal statistical analysis. I can't see how observing someone doesn't present evidence, and making a value judgement on that is statistical analysis.
Again, you seem to have a very narrow view of statistics. The fact that you began with “value judgment” is enough to show it is statistical. Value refers to some relative amount on a scale, and that’s handled statistically.
In this case, why do you think someone who presents no evidence for a claim is doing a bad job? Because in your extensive life experience, you have found that claims are usually supported by evidence. So you are in effect constructing a scale—amount of evidence supporting a claim—and putting that person at one end of it. I’m sure if he presented some weak evidence, you would have a better, but still not really good, view of him. If he had presented strong evidence, you would have a much better view of him.
This doesn’t mean that your evaluation of the person is linearly related to the amount or strength of evidence he presents. But it surely is related, in the sense that the more or stronger the evidence, the better the job in your view.
consequence of guilty verdict (you're not omniscient ;-)) - no I'm not omniscient, but I did know, as an absolute fact, that a guilty verdict would have meant a custodial sentence as defined by statute (the probability of that was literally 1, so no 'internal estimation' was needed). I just didn't think that was 'fair', based on my own sense of morality and ethics; depriving someone of their liberty is a pretty 'big' decision, and ultimately, I took the view that even if the guy probably was definitely guilty, the very fact of sending him to prison (in and of itself) would have been wrong in the abstract, and I didn't want to do that.
Our legal system does make a simple, binary distinction: guilty or not guilty. But that doesn’t mean that we determine guilt in this fashion. We determine the degree that we believe there is guilt, and somewhere on this scale, we find it’s enough for commitment to a verdict of guilt. People may differ where they find the degree sufficient—just as CAS allows arbiters to find guilt anywhere between preponderance and beyond reasonable doubt—but the fact that there is variation in where the line is drawn (which in itself is probably determined in large part statistically), doesn’t mean that the framework in which a line is drawn is not constructed statistically.
If you saw someone commit a sadistic, brutal murder, no doubt you would have no problem thinking it fair to convict that person to life in prison, or perhaps death. Why? Because on the scale of evidence, witnessing a crime is essentially 100%. That’s as sure as you can be. But if your evidence is somewhat lower on the scale, you may not consider that verdict fair. Why? Because, whether you realize this or not, you have a concept of what is fair based on your degree of certainty. This is statistical reasoning.
Look, I'm definitely not disagreeing that internal statistical analyses happen literally all the time to inform the judgements we make about the world. But inform is the key word there - sometimes (as in my case as a jurist) humans handwave the data because it doesn't match with the attitudinal judgements we're making based on our own internal value systems.
And as I’ve been trying to explain, our internal value systems operate statistically. What you call handwaving is just individual variation in where on the scale of evidence guilt is set. There is nothing whatsoever about this that is incompatible with statistics. You seem to think that statistics is supposed to provide a definite answer, the same for everyone. I think some others are arguing on this basis. But that is not what I mean when I say that we reason statistically. Of course the line is drawn somewhat differently by everyone. In this sense, statistics is value free. Statistics tells us what the probability of something is, but individuals vary in what probability is high enough for them to assign guilt.
RobbieCanuck said:
You see, you have proved my point. The median is simply another indicator of central tendency. The emphasis here is on the words "indicator" and "tendency." The median simply divides the distribution of the data set into two parts such that demonstrate an equal number of scores above and below the median.
To repeat, you’re greatly oversimplifying statistics. Human beings have a need often to make a simple binary distinction, as between guilty or not guilty. But statistics is not by any means confined to doing this. Statistics is completely compatible with grades of guilt, as in preponderance vs. beyond reasonable doubt.
It puts a different context on the data set than does the mean. Therefore we now have two statistics that are indicators of central tendency but have totally different meanings. This is why decision making and problem solving cannot be reduced to a statistical analysis
On the contrary, it’s humans who demand different kinds of meanings, and find them in different interpretations of statistics. I think you also are assuming that statistics is supposed to provide one definite answer for everyone. Of course, people interpret statistics differently, that doesn’t mean everyone doesn’t make use of them.
hrotha said:
Huh? People do that all the time. It's the basis of market research, PR, political campaigns, huge etcetera.
Not to mention our economic system, our approach to climate change, our decisions about going to war, how best to help other countries. There isn’t any major political or social decision I’m aware of that isn’t made in large part by reference to statistics. At the social level, this is widely acknowledged. My point is that this occurs at the individual decision level as well, but people are not as aware of it.