On the nature of evidence

[cheesehead]

I've noticed that some folks seem to have trouble with the concept of evidence.

In view of Brian's recent reminder (http://www.mersenneforum.org/showthr...485#post342485) that there's a difference between discourse and dissertation, I'm posting my evidentiary remarks in this new thread, rather than clutter the other.

I'm gradually learning some of the fine points of distinguishing between what is claimed to be evidence and what actually is evidence. I'll illustrate with a couple of stories, then explore the role of evidence in some ... situations of local interest.

- - -

Forty years ago I was a systems programmer at the Tulsa research center of a major oil company. We had an IBM 1800 mainframe. IBM's main advertising thrust for its 1800 was that it made an excellent process control computer. Some knowledgeable research center folks, plus the IBM customer engineers (CEs), had noticed that the 1800 was capable of being more than "just" a process control computer. Indeed, it was capable of doing as much of the research center workload as a much more expensive IBM 360 system -- after a few years of programming tweaks to the 1800's operating system.

One of my projects was to write the I/O driver, test program, and application software for a digital drum plotter designed to operate on an IBM 360 selector channel. Though a selector channel was not part of standard 1800 equipment, IBM made it possible to add such a channel to the 1800 as a special feature, which we had.

So, I wrote the software. Because of data rates and 1800 memory size limitation, the plotter application had to simultaneously feed data to the plotter while reading data for the next plotter row from tape. As a consequence, I was pleased to write my (a) first-ever I/O driver, (b) first-ever double-buffering routine, and (c) first-ever selector channel program. :-D

The plotter was photographic. It had a cylindrical drum on which a large sheet of photosensitive paper was unrolled from a tube and exposed to light flashing digital data from a CRT as the drum rotated.

All test patterns printed perfectly. But when we ran the application that read seismic data from a tape and wrote it to the plotter, there would sometimes appear seemingly-random "static" data points on some parts of the plot instead of what was supposed to be there.

-

My boss decided that the coincidence of timing was solid "evidence" that there must be some bug in my software. After all:

1) The plotter had previously been successfully used for years connected to a 360 system elsewhere in the company.

2) No "static" had ever appeared then.

3) The selector channel was the same as used on the 360.

4) We had disk drives running successfully on that selector channel on the 1800.

5) All plotter manual diagnostics were fine.

So, the hardware associated with the plotter and selector channel seemed all okay (... but the plotter and selector channel weren't the only hardware involved ...)

-

What else could be different besides my software versus the software used on the 360 system? After all, there had been a lot of "firsts" for me in writing this software. Surely the problem lay there.

I looked hard, but saw no way for my software to send "static" to the plotter.

I thought there might be something else different, something not obvious, but what in the heck could it be?

-

Our CE suggested paring the situation down to the simplest possible hardware/software involvement. We arranged to come in Saturday morning to use his diagnostic equipment and skills to help track down the problem.

We manually entered (through bit switches) a very basic program (a couple of instructions) to write data to the plotter. Nothing went wrong. Then we manually entered a very basic program to read data from the tape. All went well.

Then we entered a combination of instructions to simultaneously read from the tape and write to the plotter. When we developed the photographic paper, there was the "static"!

The CE thought there might have been some wiring mistake when the plotter had been added to the 1800. He checked the wires for the plotter. All OK.

Then we both realized that the plotter wasn't the only suspect. This had been the first time ever that the 1800 had used plotter _and tape_ simultaneously. So he checked the wires for the tape drive.

Bingo!

The tape drive had been mistakenly connected, years earlier, to the wrong cycle-steal priority, the priority that was supposed to be reserved for the plotter. (Cycle-steal priority determined which I/O device got first dibs on connection to memory for transferring a data byte.)

That mistake had never caused any trouble before the plotter was connected. But now that both were connected, use of the same cycle-steal priority for both meant that sometimes data bits would flow directly from tape to plotter -- not what's supposed to happen! (The tape data was supposed to be properly arranged in memory by software before being transmitted to the plotter.) Result: "static" whenever the I/O cycle-steals of tape and plotter happened to coincide.

Our CE fixed the tape wiring, and all "static" disappeared from plots.

The coincidence of timing of (a) first use of my software and (b) "static" appearance turned out not to be actual evidence.

The _causal_ coincidence was really (a) first overlap of tape drive reading with other I/O on the same cycle-steal priority and (b) "static" appearance.

The new plotter's use revealed an old non-plotter problem that had been lurking unnoticed for multiple years.

(First story's end)

- - -

(For the record: this thread's original title was "On the nature of evidence". If you see a different title now, it may be an improvement.

I'm fond of some title styles common in the 18th and 19th centuries, but maybe not everyone else here is.)

[Brian-E]

Reading your story about the plotter and tape drive, I do have some sympathy with your boss' interpretation of the evidence as was available to him/her at the moment that your software was first tried with the seismic data, even though you were later able to show that your boss' interpretation was incorrect. As someone who has in the past been in charge of supervising programmers myself, my initial response, on seeing a new piece of software fail to run correctly on hardware which had never before shown any defects of the kind, was most certainly always to tell the programmer to go back and find the bug in what he or she had written. Then again, I was not always as open minded as I should have been.

I guess your account illustrates very nicely that evidence will be reasonably interpreted differently by one person than by another. Your boss will have judged the explanation of your having made an error to be a much more likely interpretation of the available evidence than incorrectly wired hardware would have been. Perhaps, after checking and double checking your work, you yourself interpreted the evidence differently - and correctly - at the time before the wiring problem was discovered?

By the way, we had an application programmer (not in my group) who became notorious for complaining that there must be a bug in the compiler or the operating system whenever his program would not work and he could not see why. ("There's a bug in UNIX!") After more than one such incident, when this programmer had to be shown his own error, such pronouncements by him were unsurprisingly treated with some disdain! The reputation of individuals involved can be an important weighting factor when considering evidence.

[cheesehead]

Originally Posted by Brian-E

I guess your account illustrates very nicely that evidence will be reasonably interpreted differently by one person than by another. Your boss will have judged the explanation of your having made an error to be a much more likely interpretation of the available evidence than incorrectly wired hardware would have been.

Thank you these comments. I needed to provide a more complete explanation of the role of evidence in this case.

No, it wasn't a matter of different interpretation of evidence; it was a matter of originally overlooking one of the clues (first-time tape/plotter overlap) as to where to look (wiring) for evidence.

No one even thought of the possibility of a wiring error at first, so my boss was -- quite understandably (I'm not blaming him!) -- never weighing two different interpretations. It seemed reasonable to everyone, at first, that the only possibility was a program bug.

Perhaps, after checking and double checking your work, you yourself interpreted the evidence differently - and correctly - at the time before the wiring problem was discovered?

Again, not a matter of different interpretation, but a matter of suspecting that the evidence was incomplete. I tried to find other ways of finding more evidence, but it was our CE's suggestion (pare to the bare minimum) that got us on the right track.

I never, ever suspected a wiring problem because I had assumed, like everyone else, that the tape's wiring must be correct because it had worked for years -- right up to our (CE and I) shared sudden realization on that Saturday morning.

It was that common (but incorrect!) assumption that all was okay with the tape drive that had kept everyone from suspecting the real cause. We had all made the assumption that the years of correct tape operation must mean that there was nothing wrong with it.

There had been insufficient double-checking at the time the tape drive was originally attached. No one then had noticed the wrong (compared to what had been specified) cycle-steal priority hookup. (Or, at least, the guy who made the wiring mistake hadn't corrected it if he did notice. Maybe he thought either priority was okay, without considering the future already-planned plotter hookup.) The correct priority was in the official written documentation all along!

By the way, we had an application programmer (not in my group) who became notorious for complaining that there must be a bug in the compiler or the operating system whenever his program would not work and he could not see why. ("There's a bug in UNIX!") After more than one such incident, when this programmer had to be shown his own error, such pronouncements by him were unsurprisingly treated with some disdain! The reputation of individuals involved can be an important weighting factor when considering evidence.

I think I noticed that my boss treated my opinions, and programs, with more respect after the plotter problem was resolved.

[cheesehead]

The remaining stories will all be shorter than the first one. Also, I won't be the evidence-hero in any of them.

Indeed, I am the evidence-dunce in the next two.

- - -

A few years later than the first story, I read one of Immanuel Velikovsky's books (http://en.wikipedia.org/wiki/Immanue..._by_Velikovsky), perhaps Worlds in Collision. I was impressed by his description of how his theory matched up with events described in ancient literary works (e.g., the Old Testament), and how his theory suggested that certain historical chronologies needed revision. (Note: I was impressed by his description, not by actually comparing the ancient literature to his theory on my own.)

By that age, I had studied enough astronomy, and in particular celestial (orbital) mechanics, so that I ought to have realized right away that Velikovsky's account of careening planets was physically impossible. There existed the evidence of hundreds of years of observation and measurement of planetary positions. This evidence had thoroughly shown that the rules governing planetary motion were well-enough known for reliable predictions to be extended thousands of years into the future and (most relevant here) into the past. Velikovsky's theory impossibly conflicted with those rules (e.g., conservation of angular momentum).

But I was in a what-if-conventional-wisdom-is-wrong stage of life. Also, at that time I'd lost interest in scientific astronomy. I let my mood allow me to be convinced by Velikovsky's dramatic writing.

Only after I read some criticism of Velikovsky's theory several years later did it dawn on me that well-established orbital mechanics evidence showed Velikovsky's theory to be impossible.

(End of second story)

[cheesehead]

Around the same time as the second story began, I read another book, The Jupiter Effect (published 1974). Its authors propose a theory that a planetary alignment in 1982 (eight years after book publication) will cause a noticeable increase in earthquakes -- not directly by their gravitational pull on Earth, but through a chain of effects:

1. The planets will be lined up on the same side of the sun during 1982. Since they orbit the sun at various distance and speeds, this event is very unusual.
2. Each of the planets exerts a gravitational tidal pull on the sun. When aligned, their combined force will substantially affect the sun.
3. The sun will respond with increased surface activity in the form of sunspots and eruptions. This also implies an increase in ejected solar wind particles (charged particles such as electrons and ions) throughout the solar system. Those impinging on the earth will cause large-scale movement of air masses.
4. The resulting atmospheric disturbance will alter the earth's rotation rate. Variations in the earth's spin will trigger regions of geologic instability, causing widespread earthquakes.

Their steps seemed individually plausible to me, but again I did not immediately compare their theory to what had already been firmly established by scientific evidence.

I even went so far as to tell a bookstore clerk that The Jupiter Effect should be displayed in the astronomy section rather than where they actually displayed it (some non-science section -- perhaps it was fiction, or astrology). I regret that now.

At some later time, something I saw about the scientific evidence concerning one of those steps made me look at the book again. I then realized that the proposed series of effects didn't match existing evidence in at least a couple of regards.

(End of third story, but there's a sequel)

[kladner]

The plotter was photographic. It had a cylindrical drum on which a large sheet of photosensitive paper was unrolled from a tube and exposed to light flashing digital data from a CRT as the drum rotated.

All test patterns printed perfectly. But when we ran the application that read seismic data from a tape and wrote it to the plotter, there would sometimes appear seemingly-random "static" data points on some parts of the plot instead of what was supposed to be there.

[OT] This is utterly tangential, but your account details a transition period in the analysis of oil exploration data from analog to digital. My Dad was involved in this field from the completely analog to the 360 mainframe era. He moved from seismograph field operator to the main office transcription (my term) operation. He was part of the crew which ran the "playback machine". It read wide magnetic tapes and plotted the data. Rather than exposing photographic material, it burned (literally, it had suction hoses to remove smoke) traces onto specialized paper with some kind of electric needles. I was pretty young, and only saw this monster a few times, so I don't have extremely clear memories of the details. I believe this was all analog. I know it ran on vacuum tubes. It was a cranky, finicky beast.[/OT]

It fascinates me that your occupation was so close to my father's. He ended up learning to write Fortran routines which ran on the above-mentioned 360's via punch cards, though tape drives and disk packs were also involved.

[cheesehead]

Sequel to story 3:

BTW that quote in story 3 was from a source I didn't find until just now as I wrote this post: Institute for Creation Research (!!!)

Back in the 1970s, when the third story took place, I was completely unaware of what ICR had said about this book.

Though the ICR has always failed to apply proper skepticism to its creationist theories, at first glance its comments on The Jupiter Effect seem to show that it should have been capable of doing so, if it were not so blinded by its theological principles.

Here's the link: http://www.icr.org/article/154/

As far as I can recall without finding and re-reading the book, ICR's description of the four steps in the chain of events proposed in The Jupiter Effect is accurate as far as it goes.

Then, ICR's following analysis (on that same page) of each link initially seems reasonable overall. (Note: I haven't explicitly fact-checked but am relying on my memory (Alert! Danger, Will Robinson!) of the astronomy involved.) Many statements there are casual rather than rigorously scientific, but not obviously incorrect.

There is a noticeable, but minor, typo in the analysis of link 2:

Originally Posted by http://www.icr.org/article/154/

Just how significant is the solar tidal force? ... Hence, the Jupiter-sun tidal force is only that of the familiar earth-moon force.

(Should have been "is only (some fraction) of the familiar earth-moon force".) But this is quickly revealed by their next sentence:

That is, the earth tidal pull due to our moon is nearly one-third million times greater than the solar pull due to Jupiter!

However, the theologically-inspired last sentence of that section

The stable sun certainly appears to have been formed for man's advantage, not for his annihilation!

has no factual justification accompanying it.

The writer of ICR's analysis of link 3 seems not have understood the evidence of ICR's own Table 1, further down! There's a serious oversight!

Excluding Pluto, a rough alignment of the planets Jupiter through Neptune occurs about every 179 years. Thus, the Jupiter Effect hypothesis may be evaluated by checking the records for the similar alignment of 1804.

No, the Jupiter Effect hypothesis may be evaluated by checking the records for alignments of the planets exerting the largest tidal effect on the Sun -- which is not at all the same as "the planets Jupiter through Neptune".

Table 1 clearly shows that the minimal tidal effect (0.54) of Mercury alone is almost five times as great as the combined tidal effects (0.1126) of Saturn, Uranus and Neptune! Mars's minimal tidal effect (0.023) is nine times the combined tidal effects (0.0026) of Uranus and Neptune.

So, if one were to check the Jupiter Effect hypothesis by looking at what happens on the Sun when the most influential planetary alignments occur, one should look at alignments of Jupiter, Venus, Mercury, and Earth before considering where Saturn, Uranus and Neptune are located.

In contrast, the writer was on-target with his next paragraph, starting with

If the Gribbin-Plagemann theory is correct, there should also be high solar activity whenever the planets are lined up on both sides of the sun.

He is correct about effects at oppositions versus effects at conjunctions.

- - -

ICR was correct in concluding

It is clear that the scientific reasoning for the Jupiter Effect is invalid.

But it was sloppy in its own scientific reasoning, even when unbent by theology.

[chalsall]

Originally Posted by cheesehead

I've noticed that some folks seem to have trouble with the concept of evidence.

Cheesehead. A sincere question...

Are you familiar with the concept of a Signal to Noise Ratio (SNR)?

[cheesehead]

Originally Posted by chalsall

Cheesehead. A sincere question...

Are you familiar with the concept of a Signal to Noise Ratio (SNR)?

There's not much signal in that post. Please explain what you mean instead of running the risk that I will misinterpret your tiny hint.

[chalsall]

Originally Posted by cheesehead

There's not much signal in that post.

Exactly.

[only_human]

My revolutionary theory does not need evidence because the predictions are imaginary and thus do not need a real test! When my theory is finally appreciated, everyone will see what a sham modern science truly is! I'd love to get those scientists to take back all that proof stuff that is polluting the evidence! They might shut me up like Galileo, even though I don't expect the Spanish Inquisition.

If the scientific [sic] establishment is caught in the suppression of my well-deserved fame, then maybe I won't be persecuted like a red shirt on Star Trek - maybe Star Wars storm troopers from that project that the government said was too costly (secret Death Star) will come but I'm not scared even though all the shrinks say I am wrong. Einstein woulda got this far if he knew about the extraterrestrials mentioned in that suppressed sequel to Contact. Feynman tried to hide it by saying that he didn't buy that string theory stuff but he knew - he knew because he could hold a glass of water in one hand while playing the bongo drums in another while rotating the water glass arm without spilling a drop. He didn't brush before going to the dentist - that is incontrovertible proof!

TL;DR, well, that's just more suppression. I have to be the defender of the old guard and other orthodoxies too numerous to mention. The hidebound reactionaries can't stop the signal. My theory explains magnets with CRTs, which is why they are getting rid of CRTs (SO NO ONE CAN FIND OUT!)

Everyone should know the Dr. Kintsugi (金継) field equation - if they could understand it! They won't dismiss it like my other theories, which were right BTW. Dune tells us what the planet will look like unless we fix Newtons stuff. Ha! Classical mechanics just need a few major fixes (especially the ones who work on old Mustangs). Too bad there is no Nobel Prize for math or I would get one for sure! That's why I am not sending it to any of those bean counters; it'd test high, I know but I am not sharing. It's a paradigm shift, like Einstein! Current theories predicted a new boson, but they don't know about Garçons or Moreons. Anyway, they are only theories. I just need someone good at math to write it down. I'll share my lottery numbers with the first person to find a flaw in it! Don't tell anyone else, I don't want these thoughts stolen :-) I did it all on my own, even went to school for a long time so you know it is good.