Hacker News new | past | comments | ask | show | jobs | submit login
Howland Will Forgery Trial (wikipedia.org)
66 points by Hooke 13 days ago | hide | past | favorite | 21 comments

While I appreciate that they attempted to use some kind of scientific reckoning, using a weak science (handwriting analysis) and then adding ridiculously high numbers to it (1/2.666E21) is a really bad way to decide on this.

Link below [2] explains that the probability was NOT 1/2.666E21 that it was not fake, because it is more likely than that that the science of handwriting analysis is incorrect, or that the analysis was done incorrectly, or anything similar (i.e. A one-in-a-trillion chance is more likely to be incorrect than one-in-a-trillion).

Bonus round: They also commit the "Prosecutor's fallacy" [1], where they explain how likely the evidence is given the hypothesis instead of explaining how likely the hypothesis is given the evidence. Example: You are found in a store and accused of robbing it. They ruled that you were accused of robbing the store, and you being in the store fits this theory, when they should have ruled on how likely you were to have robbed a store when they only knew you were in it (not very likely, lots of people may have been in the store).


> using a weak science (handwriting analysis) and then adding ridiculously high numbers to it (1/2.666E21) is a really bad way to decide on this.

Ah, but ridiculously high numbers get such a nice emotional reaction from a jury.

I once had a conversation with a guy who fiercely proclaimed that if you shuffle a deck of cards, the resulting ordering is likely to be utterly unique in the history of all shuffled decks ever, because there are <dramatic pause /> 8E67 possibilities!1! Same as the number of atoms in our galazxy!

I went on to explain that while in theory this is true, in practice there are a lot of caveats. Since brand new unopened deck of cards are usually all in the same order, and since a perfect cut and a perfect riffle shuffle (left/right/left/right/left etc) will always produce one of two orderings, depending on whether you start with the left or right decks, etc etc, it lowers the odds drastically. And it would be more proper to say that if you sloppily shuffle a deck about 7 times, then the original claim is probably true.

But I lost the crowd. They had already had an emotional reaction to the enormous number .

How is it possible that a wiki page about a trial neglects to mention the outcome?

> The Circuit Court finally refused to admit the complainant's testimony, and dismissed the bill with costs. An appeal was taken, but withdrawn, on a settlement between the parties whereby the complainant received her expenses, costs, and counsel fees. The probate of the will of September I, 1863, therefore, remained undisturbed,

From https://scholarship.law.upenn.edu/cgi/viewcontent.cgi?refere...

Which references as the original source:

> The American Law Review (vol. 4, pages 629-54)

Not sure if this was added before or after your comments, but at the time I'm looking, the last scene sentence in the second to last paragraph of the intro is "The case was ultimately decided against Robinson after the court ruled that the clause invalidating future wills and Sylvia's signature to it were forgeries"

Looking at the edit history it seems that someone came in and gave the article a general cleaning.

I wonder if it's a viable strategy to post sub-par, but interesting pages to HN in order to attract high-quality edits...

A one in 2,666,000,000,000,000,000,000 chance that 2 signatures would match... Sounds about right for my own signatures ;-) When signing multiple documents I get very different versions each time. As in.. some expert would probably think they’d be fakes.

Are signatures still this valuable in court, in this day and age? Given that they should be easy enough to properly forge, instead of straight out copy.

I used to get stressed at supermarket checkouts back when you had to sign your name on the credit card slip. More than once I was asked to do it again and then you felt all eyes on you and it was even harder.

The last time I did mortgage paperwork, I signed my name over and over so many times that I had to start taking care that there weren't visible differences between the signatures.

Signatures are bullshit. And the line:

> the probability that all 30 downstrokes should coincide in two genuine signatures was 1 2.666 × 10 21 {\displaystyle \textstyle {\frac {1}{2.666\times 10^{21}}}} \textstyle\frac{1}{2.666 \times 10^{21}}. That is one in 2,666,000,000,000,000,000,000, in the order of magnitude of sextillions.

is such outrageous bullshit that I'm saddened he wasn't immediately laughed out of court.

Outrageous bullshit like this appears in modern court cases all the time today. For example, it is common to cite astronomically small odds that a DNA match is in error, completely ignoring the far higher odds that the sample was accidentally contaminated, or the lab deliberately falsified the result to satisfy prosecutors, etc. I'm not saying that DNA evidence is bullshit, only that the only way to get something like a 10^-21 odds of error is to ignore all the largest possible error sources.

So one way of determining if an instance of a signature is genuine is that it should differ to a significant degree with all other known previous instances. What a brilliant use of probability so long ago

Similar to will forgery, deed forgery is still a problem today. In fact, it is a growing problem, as forgers have the information and access they need through online land records. While online land records have greatly simplified real estate closings, this is one downside. County recorders are unable to stop these forged deeds, as they are legally required to record all deeds that are appropriately formatted.

Many counties have deed recording alerts that alert property owners to documents recorded against their name or property. However, many counties do not offer this service, leaving property owners unaware of these forged or fraudulent deeds, until they receive an eviction or foreclosure notice. As such, we are offering a Title Watch service to inform you about potential forged or fraudulent deeds recorded in your name or against your property. Please visit us at https://plathq.com

(1) Interesting connection to the original story. This is the sort of not-exactly-an-ad-because-it-is-related mention that I find cool about Hacker News.

(2) Let me see if I get this right: you offer a service that simply monitors deeds and mortgages recorded against a specific plat and sends an alert to to the subscriber if something is recorded (which is an extremely rare event). And you charge $10/month/property for this service? What about the problem makes it so expensive? Why aren't you likely to be disrupted by someone like Zillow offering this $10/month service for FREE (and making money off the address list they develop from it, plus advertising to all the homeowners)?

Glad you liked the “not-exactly-an-ad-because-it-is-related mention.” While the probability of a forged deed is relatively low, the impact is not. In addition to the risk of a thief just forging a deed to a property they have no connection to, there is also the risk of intra-family deed/mortgage forgery. This intra-family forgery is likely substantially underreported, and causes the family to lose the property it has acquired through hard work because of one bad apple.

In addition, people are (rightly) highly risk-averse about anything that affects their family, their home, or the assets that they built over years and decades. For example, real estate investors commonly put their property into an LLC (which costs $800/year in CA) or buy liability insurance, even though the probability of incurring a large liability is low.

With respect to costs, even though land records are public, machine-parseable title/recorder data is expensive. Data from the tax assessor (even machine-parseable) is relatively cheap, but is not as complete as recorder data. For example, it does not have mortgages. We intend to achieve economies-of-scale with respect to recorder data. At that point we intend to offer additional value at the same price (for example, insurance) and add a free-but-sell-leads service.

The Peirce family are some of America’s most under appreciated mathematicians and logicians.

It must be supremely frustrating to a quantitative person when the statistics meet reality: a jury that says, "well, there's still a possibility he didn't do it, isn't there? You want to have us convict someone on math?"

The standards in question are preponderance of the evidence (civil) and beyond a reasonable doubt (criminal). The standard is never "any possible way" or "no possible way."

So my response would be that at a certain point, a tiny possibility does not amount to reasonable doubt.

Many people, especially when you have average people on a jury, are not very good at drawing lines where the certain % or probability equals less than a reasonable doubt.

When I read this quote, I think that either - we have lost a great deal of literary education - or we have lost a lot of hogwash.

  "So vast improbability is practically an impossibility.
  Such evanescent shadows of probability cannot belong to
  actual life. They are unimaginably less than those least
  things which the law cares not for. ... The coincidence
  which has occurred here must have had its origin in an
  intention to produce it. It is utterly repugnant to sound
  reason to attribute this coincidence to any cause but

> The coincidence which has occurred here must have had its origin in an intention to produce it.

That is awesome!

But did they find out if she was a witch?

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact