Will Scotland Yards HOLMES 2 and AI solve Jack the Ripper?

[seanr]

It truly amazes me in the year 2025 how technology averse so many on this Forum truly are. If you say NO then please defend your conclusion. What do you mean by NO?

It irritates me that realistic expectations of what these technologies are capable of, is somehow being presented as technology averse. I work with AI for my day job and I am familiar with how it works.

I say ‘No’ for the simple reason is there is way too much money being made on the case for it being unsolved. Even if the actual perpetrator of these crimes has been named already amongst the very many who have been accused, there isn’t going to be a confession or a case built which will persuade the CPS to proceed with a charge. There will be no trial and without a trial, there will be no case which has been tested before a jury and proven beyond reasonable doubt. So, there will always be doubt and in our legal system the accused is innocent until proven guilty in a Court of Law. So, it can never be solved. Technology cannot change that.

Any reasonable doubt at all, will always be enough room for someone to put forward their new theory, like maybe the Tetley Tea Folk or the Mob did it (or perhaps the Tetley Tea Folk in cahoots with the Mob - and the Royal Family), and get a book deal and fawning Daily Mail coverage. Whilst financial rewards for providing a new names and theories remains high, then new suspects will continue to be a production line of spurious reasoning.

You seem to place objective data on some kind of pedestal which makes it somehow incontrovertibly reliable but that’s not how data works at all. Data is not knowledge, data is the raw material from which we attempt to extract or build knowledge or failing that we make decisions with it.

To give a tangible example, when we use machine learning in an online payment for assessing whether a particular transaction may be fraudulent we may look at characteristics of the transaction and compare them with known attributes of fraudulent transactions, things like time of day, amount, credit card supplier, customer name, how familiar the customer is to this provider, whether the transaction is an outlier compared to usual customer behaviour… and so on. The machine learning can return us a score, say 90% probability that a transaction is fraudulent. And we may, to protect our platform from financial loss, decline all transactions where the probability of fraud is thought to be above 90%.

This does not mean, and must not be taken to mean, the transaction has been proven by technology to be fraudulent. At a certainty of 90% (if the probability is accurate), for every 90 fraudulent transactions prevented by such a system, 10 legitimate transactions would be blocked from completing. This also means fraud is not proved beyond a reasonable doubt, this evidence should not be used to take the customer and deprive them of their freedom for the crime of fraud. It would be an error in criminal law to conclude that the machine says it’s 90% likely the suspect is a criminal, therefore they must be guilty.

The technology cannot prove the case. Building a proper legal case still requires proper police work, piecing together the evidence and considering the negative case, that no crime was attempted by the suspected person at all. This should result in proper caution about how such technologies can and should be used in criminal investigations. They are not the answer machines, they can be supportive technologies guiding where to spend time and resources to build a case or find a suspect or person of interest.

In the case of using the technology for the Whitechapel Murders, it would be necessary to collect more information from an outcome, putting proper boots on the ground in Whitechapel, proving motive, means and opportunity. Making an arrest, interviewing the suspect… It’s all obviously too late to do that work.

Without the accompanying boots on the ground, the technology can only ever be advisory. At absolute best when applied to the Ripper case, it can provide interesting hints at possibilities but at worst, powerful tools for assisting live investigations is perhaps wasted on little more than an entertainment, a sensationalist media parlour game.

You seem determined to treat eyewitness testimony as though it is hard data, going so far as to suggest that if it is inaccurate the eyewitness may be subject to perjury charges. The standard for perjury is, thankfully much higher than this. Study after study has shown that recall and memory can be inaccurate and influenced. I can point you to multiple examples of H Division’s treatment of eyewitnesses in the late 19th and early 20th century not being up to the 21st century professional standard, they were at heightened risk of influencing witnesses compared to modern police.

A widely remarked upon stat from the Innocence Project says that “Eyewitness misidentification is the single greatest cause of wrongful convictions nationwide [in the USA], playing a role in 72% of convictions overturned through DNA testing”. In the UK, a man was convicted after being identified by 4 eyewitnesses and was later proved to be innocent. Source: https://www5.open.ac.uk/research-cen...ness-testimony

None of the eyewitnesses in these cases seem to have intended to mislead, they were in error and so were not guilty of perjury.

Betteridge's Law of Headlines often works for forum thread titles. So, I say ‘No’ and I am quite certain that ‘No’ is the correct answer. Count yourself lucky I do not answer ‘LOL, NO!’.

[John Wheat]

It irritates me that realistic expectations of what these technologies are capable of, is somehow being presented as technology averse. I work with AI for my day job and I am familiar with how it works.

I say ‘No’ for the simple reason is there is way too much money being made on the case for it being unsolved. Even if the actual perpetrator of these crimes has been named already amongst the very many who have been accused, there isn’t going to be a confession or a case built which will persuade the CPS to proceed with a charge. There will be no trial and without a trial, there will be no case which has been tested before a jury and proven beyond reasonable doubt. So, there will always be doubt and in our legal system the accused is innocent until proven guilty in a Court of Law. So, it can never be solved. Technology cannot change that.

Any reasonable doubt at all, will always be enough room for someone to put forward their new theory, like maybe the Tetley Tea Folk or the Mob did it (or perhaps the Tetley Tea Folk in cahoots with the Mob - and the Royal Family), and get a book deal and fawning Daily Mail coverage. Whilst financial rewards for providing a new names and theories remains high, then new suspects will continue to be a production line of spurious reasoning.

You seem to place objective data on some kind of pedestal which makes it somehow incontrovertibly reliable but that’s not how data works at all. Data is not knowledge, data is the raw material from which we attempt to extract or build knowledge or failing that we make decisions with it.

To give a tangible example, when we use machine learning in an online payment for assessing whether a particular transaction may be fraudulent we may look at characteristics of the transaction and compare them with known attributes of fraudulent transactions, things like time of day, amount, credit card supplier, customer name, how familiar the customer is to this provider, whether the transaction is an outlier compared to usual customer behaviour… and so on. The machine learning can return us a score, say 90% probability that a transaction is fraudulent. And we may, to protect our platform from financial loss, decline all transactions where the probability of fraud is thought to be above 90%.

This does not mean, and must not be taken to mean, the transaction has been proven by technology to be fraudulent. At a certainty of 90% (if the probability is accurate), for every 90 fraudulent transactions prevented by such a system, 10 legitimate transactions would be blocked from completing. This also means fraud is not proved beyond a reasonable doubt, this evidence should not be used to take the customer and deprive them of their freedom for the crime of fraud. It would be an error in criminal law to conclude that the machine says it’s 90% likely the suspect is a criminal, therefore they must be guilty.

The technology cannot prove the case. Building a proper legal case still requires proper police work, piecing together the evidence and considering the negative case, that no crime was attempted by the suspected person at all. This should result in proper caution about how such technologies can and should be used in criminal investigations. They are not the answer machines, they can be supportive technologies guiding where to spend time and resources to build a case or find a suspect or person of interest.

In the case of using the technology for the Whitechapel Murders, it would be necessary to collect more information from an outcome, putting proper boots on the ground in Whitechapel, proving motive, means and opportunity. Making an arrest, interviewing the suspect… It’s all obviously too late to do that work.

Without the accompanying boots on the ground, the technology can only ever be advisory. At absolute best when applied to the Ripper case, it can provide interesting hints at possibilities but at worst, powerful tools for assisting live investigations is perhaps wasted on little more than an entertainment, a sensationalist media parlour game.

You seem determined to treat eyewitness testimony as though it is hard data, going so far as to suggest that if it is inaccurate the eyewitness may be subject to perjury charges. The standard for perjury is, thankfully much higher than this. Study after study has shown that recall and memory can be inaccurate and influenced. I can point you to multiple examples of H Division’s treatment of eyewitnesses in the late 19th and early 20th century not being up to the 21st century professional standard, they were at heightened risk of influencing witnesses compared to modern police.

A widely remarked upon stat from the Innocence Project says that “Eyewitness misidentification is the single greatest cause of wrongful convictions nationwide [in the USA], playing a role in 72% of convictions overturned through DNA testing”. In the UK, a man was convicted after being identified by 4 eyewitnesses and was later proved to be innocent. Source: https://www5.open.ac.uk/research-cen...ness-testimony

None of the eyewitnesses in these cases seem to have intended to mislead, they were in error and so were not guilty of perjury.

Betteridge's Law of Headlines often works for forum thread titles. So, I say ‘No’ and I am quite certain that ‘No’ is the correct answer. Count yourself lucky I do not answer ‘LOL, NO!’.

Great post seanr. A.I. is currently by its very nature unreliable throw in witnesses which mat not have seen Jack the Ripper and the fact that witness testimony is notoriously unreliable and this study is unreliable and extremely limited.

Cheers John

[Patrick Differ]

Seanr..the comment was made there was no objective ( factual) data in this case. It's not a pedestal.I was drawing a comparison to fiction.

I can't help you with the irritation but since you understand Big Data analytics and Automation ( the world i come from ), then you know it has benefits. There is also real-time engineering data which is more up my alley.

Did not mean to generate a hornets nest but I personally thought that the HOLMES2 Platform is useful in the same manner that Steve Bloomers maps in Ripperologist 173 are useful. Gets you thinking about possibilities. Helpful not harmful.

So this small group says NO to HOLMES and AI and we turn over more rocks. That's fine..

[Fiver]

Fiver you are aware that perjury was a punishable crime in 1888 London? Imprisonment.

You are aware that nobody on this thread has accused any of the witnesses of perjury?

You are aware that sometimes the witnesses contradict each other?

Lawende, Levy, and Harris saw a man and a woman near Miter Square.

* The man could not have been 5'3", 5'7", and 5'9" tall.
* The man could not have be "of shabby appearance" and be "dressed respectably".
* The man could not have had both fair and brown hair.
* The man could not have had both a big and a small mustache.
* No witness could have seen the size and color of the mustache if the man's back was turned.
* No witness could have seen what the man was wearing around his neck if the man's back was turned.

The Levy account contradicts Lawende. The Harris account contradicts Lawende. The Lawende accounts contradict each other.

For the third time - "Witness accounts can be inaccurate without anyone lying. Human perception and memory are fallible."

That's before we consider that we don't have written, signed, accounts from Lawende, Levy, and Harris. We have newspaper accounts of unknown accuracy. We have police reports that appear to have been written from memory as they repeatedly contradict each other about what Lawende said.

Human perception and memory are fallible.

[seanr]

Seanr..the comment was made there was no objective ( factual) data in this case. It's not a pedestal.I was drawing a comparison to fiction.

I can't help you with the irritation but since you understand Big Data analytics and Automation ( the world i come from ), then you know yit has benefits. There is also real-time engineering data which is more up my alley.

Did not mean to generate a hornets nest but I personally thought that the HOLMES2 Platform is useful in the same manner that Steve Bloomers maps in Ripperologist 173 are useful. Gets you thinking about possibilities. Helpful not harmful.

So this small group says NO to HOLMES and AI and we turn over more rocks. That's fine..

The question asked in the title was ‘Will Scotland Yards HOLMES 2 and AI solve Jack the Ripper?’

I’m convinced that the answer to that question is ‘no’. That’s not the same thing as saying that applying HOLMES 2 and forensic AI isn’t helpful or interesting.

[Patrick Differ]

Respectively disagree. Not interested in constant wrangling. Thats all it is.

[John Wheat]

Respectively disagree. Not interested in constant wrangling. Thats all it is.

No other posters are talking about the limits of the study and quite rightly too.

[Patrick Differ]

Duly noted. Cheers.

[Fiver]

So the man seen with Catherine Eddowes at 1:35 am by 3 men was not the man who in 10 minutes time lured her quietly into Mitre Square, brutally mutilated her and escaped as Eddowes was found by PC Watkins at 1:45 am +/- 1 minute?

Is it therefore a low or high probability it was the killer? Or was the killer waiting in Mitre Square as the other man moved on? Or were all 3 men ( not 1) mistaken?

* Three men - Lawende, Levy, and Harris - saw "a man and woman together at the corner of Church-passage, Duke-street, leading to Mitre-square".
* Lawende was the only one to claim that the woman was Catherine Eddowes.
* None of the three saw the couple enter Church Passage.
* St James Passage was another route to and from Mitre Square. A night watchman, Blenkinsop, saw people go through it, but paid no real attention to them.
* Mitre Street was the third possible route to and from Mitre Square. If there witnesses who observed it, they did not come forward.

If Lawende was correct that the woman he saw was Eddowes, then the man that Lawende saw was almost certainly her killer. If Lawende misidentified the woman, then Eddowes and her killer entered through one of the other two routes and may have already been in Mitre Square at 1:35am.

So how good of an observer was Lawende?
* Levy contradicts Lawende's description of the man, putting him 4" to 6" shorter.
* Harris claims the man's back was turned, which means it would have been impossible for Lawende to see the man's mustache or kerchief.

There are at least three versions of Lawende's description. All are secondhand. They contradict each other about the suspect's height, hair color, mustache size, and how well they were dressed. Lawende is long since deceased, so this cannot be clarified.

And that's before we consider that the couple were strangers seen for only a few seconds in very poor lighting.

[Fiver]

Objective data is Fact, not Fiction. The results from Scotland Yard, using Objective data came to a set of conclusions. To restate

1. JtR was a Marauder not a Commuter. He lived there.
2. His boundary of operation , N, S, E, W. Are defined by the victims.
3. Tabrum was the first victim as defined by Overkill as demonstrated by her wounds and as seen in successive, escalated mutilations.
4. Based on witness testimony the killer was likely Jewish.
5. Heat maps based on data- primarily witness testimony, location of murders and post mortem records associated with Tabrum, Chapman, Eddowes and Kelly show a comfort zone concentration North of Butchers Row , Mitre Street to the West Hanbury Street to the North and George Yard to the East. 4 of the 6 murders occured in this Zone and 5 of the 6 known witnesses were also from this Zone. ( Zones are used in modeling software).
6. The 4 murders defining the NSEW boundary were equidistant to Tabrums murder.

1. I cannot comment on this without details as to how they decided JTR was a Commuter.
2. What is even meant by "boundary of operation"?
3. Tabram was probably a Ripper victim. This is not a new conclusion. She might have been the first victim.
4. This conclusion is complete nonsense. We don't know if any of the witnesses saw the Ripper. They are descriptions of strangers in poor lighting. They contradict each other. None point towards a Jewish killer.
5a. Geoprofiling is a tool to try to make searches more efficient. It cannot tell where the killer lived.
5b. Where the witnesses lived is irrelevant to geoprofiling.
5c. Witnesses are Long (Chapman); Marshall, Smith, Brown, Packer, and Schwartz (Stride); Lawende, Levy, and Harris (Eddowes); Cox, Lewis, and Hutchinson (Kelly). That's 12 witnesses, not 6.
6. Any map will show that the Nichols murder was a lot further from Tabram murder, and the idea that the killings were equidistant from the Tabram killing is utter nonsense.

[Mark J D]

... there are only 2 types of Serial killers in terms of location. Marauder or Commuter. They either live in the boundary or outside of it. What would be the third?

Lechmere would. Lives outside the boundary, but for real-world reasons connected with home, family and work is actually confined within it at certain times, and with very limited degrees of freedom.

Applying a silly binary model to this real-world life generates the wrong answer of a Marauder based near Tabram's murder site. Why even bother with it?

M.