[seppi85]

Dear RAC forum

I'm a student and was considering some nerdy computer project in which a system decides on given information and its own knowledge base which party is at fault in a car accident.

My question being, how do say the claim handlers at RAC insurance Co. lay blaim? I know some obvouse rules such if right of way, getting involved in pile up last car is usally one to blaim etc etc.

But am not after instances but rather how they approach the issue do Claim Handers read off some peice of legislation? is their a handbook to which they look up?

I guess what Im looking for a set of Rules Methodlogy in establishin blaim in an accident, I've skimmed through the Road Traffic Act and to be honest havent found anything like:

If Car A gets plowed by Car B from the Right Hand Side on Roundbout Car A is at fault.

Hope this all made sense, and perhaps someone from the industry could give me some much sought after advice.

Best Wishes
Seppi

[Rolebama]

All of those which I have met use a mix of experience and common sense. There is obviously room for human error there, in the same way that there is usually a little bias depending on who they are working for. Type and location of damage relevant to road positioning and layout are used, as are witness statements, and the statements of those involved, along with any available CCTV footage, but in the end, it comes down to what seems to be the more likely cause.

[seppi85]

Dear Rolebama

Thanks for your reply but its not what really what Im trying to get that.

I'm trying to picture, what sort of books/sources claim handlers use in deciding accident claims.

Imagine it being your first day on the job as claim handler for RAC.
And details of a claim come on to your desk.
Both Parties agree on what happened 'Car A ploughed into Car B, from the Right Hand Side, when Car B Pulled out'

Where would a rookie begin, or look for reference for say a 'Accidents in roundabout'.

I'm presuming all these insurance claim handlers base their judgements on some sort of formal predicate (So be it that most of it is common sense)- that holds in court and is easily explained.

Best Wishes
Seppi

[Rolebama]

My initial answer still holds true, although your question is a very valid one. . Unfortunately, the way insurance companies work makes it a very difficult question to answer. Although the insurers themselves are responsible for adhering to the law regarding motor insurance, they do have enough leeway to settle disputes without directly involving motoring law. (This may be useful: http://search.opsi.gov.uk/search?q=m...psi_collection)
In making a decision, insurers will take into account any Police reports and Court decisions and will hypothesise about possible outcomes if one or more motorists were prosecuted. (Although if there is a prosecution, and one motorist is found guilty of a motoring offence relevant to a collision, that person is usually deemed to be 100% at fault. Regardless of whether it is a State or private prosecution)
In that sense, they work off a knowledge of Road Traffic Act, content of Highway Code, and any relevant motoring laws that could be bought into play appertaining to the individual case*.A few examples: Road Traffic Regulation Act, Road Vehicle Lighting Regulations, Road Vehicle (Construction and Use) and Road Vehicle (Approval) Regulations.
*You can find a number of relevant laws here. http://www.opsi.gov.uk/
You may have to search each individual Act by name, as there are so many motoring related ones.

I can only say that I hope this helps.

[smudger]

Aye! although they say witness accounts are taken into consideration as well as both drivers involved accounts.

Yet how many times have we read in here, where the other driver lies in order to cover him/her self, and put at least half the blame on the other driver, who was totally innocent:mad:

[seppi85]

All valid and much appreciated to the replies but assuming we live in an ideal world where both parties agree to same story.

What formal procedures do Claim Handlers use in finding out fault, is their a guidebook? a set of rules set in stone?


The reason i'm so keen on this is becouse I'm from a computer science background and wish to propose an system which given some inputs, uses an inference engine to communicate to some natural english rule base and arrive at and outcome and explain why it has made this decision.


Thanks Seppi

[Snowball]

Are you saying that you propose to re-write the rules already used by the insurance industry? That's a big step, and I think it would fall at the first hurdle, unless you came up with something new to the industry, and which would be financially beneficial to the industry.

[Rolebama]

Very simply, no. The only aspect of a collision which is carved in stone is motoring law, and for a collision to have taken place, one or both parties concerned must have been in breach of one or more of those laws. In the event of a prosecution of one driver, he is usually held to be 100% at fault. If neither, or both, are prosecuted, then, in the absence of witnesses, it may be hypothesized about who most likely caused the collision. Unfortunately, as Smudger says, people tend to lie* in these circumstances, which may well 'muddy the waters' and make it difficult for an assessor to apportion blame fairly, but a decision still has to be reached, and when there is uncertainty about who is at fault, then blame is likely to apportioned on a 50/50 basis.
*People do not necessarily deliberately lie; they could be concussed, confused or in shock, and tell what they believe to be the truth.

[seppi85]

Thanks for your replies

Well I guess the objective of any good system would be to automate and remove the unremarkable routine cases that occupy the time of the expert.

This would hopefully allow the Expert Claim Handler to deal with harder less routine cases that really does require experience and institution.

I was proposing a system 'student project' that holds in its own knowledge base the rules that govern whether who’s at fault in a car crash.

Say a routine case comes onto a desk of a claim handler:

"Car A ploughed into Car B, from the Right Hand Side, when Car B Pulled out".

Assuming both parties agree, both parties have insurance, both are licensed etc. etc.

This would be classified as routine case, to which the case requires no real time of the expert user.

The system would look up its knowledge base and identify that

Car B is a fault - due to Rule 123.56 of RAC Handbook on “who’s to blameâ€Â

I’m after this ‘handbook’ or the methods to which claim handlers go as I said in my previous post Where would a rookie claim handler begin, or look for reference for say a 'Accidents in roundabout'.

I'm presuming all these insurance claim handlers base their judgements on some sort of formal predicate (So be it that most of it is common sense)- that holds in court and is easily explained.

Seasonal Greetings!
Seppi

[wagolynn]

There is already accident investigation software available. This software automates the process of data gathering and is designed to be used at the scene. It uses information on the individual vehicles (from the manufacturer) and the laws of physics its inputs can be any or all available crash scene data. Photo graphs - for vehicle deformation and positions -, measured distances, GPS plots – for road layout, position of skid marks vehicle positions etc - . From this data, it computes backwards from the scene, to all the most probable starting points giving a probability figure for each one. In my opinion, this should be deployed at every crash scene, not so much as to apportion blame but to understand what really happened.

[Snowball]

Originally Posted by seppi85

Say a routine case comes onto a desk of a claim handler:

"Car A ploughed into Car B, from the Right Hand Side, when Car B Pulled out".

Seppi85, I am sorry, but if you consider this to be an example of a 'routine claim', then I am afraid you are way down the league in your ability to define the role of claim handlers.

That one collision can have as many variations surrounding cause and responsibility as there are claim handlers in the industry.
Taking into account of all types of weather/road conditions, along with differing patterns of driving, I can think of around a hundred different scenarios with degrees of responsibility varying from 50/50 to 100%.
And there are probably more I haven't even thought of.

[Rolebama]

Wagolynn, I really don't see how that software can work, and I would certainly challenge it in the event it was ever used against me. Who wrote the program, and where did they get their data? (Rhetorical). As I stated in an earlier post, all impact info colllected from initiial crash testing has been proved worthless in the real world.

[wagolynn]

Originally Posted by Rolebama

Wagolynn, I really don't see how that software can work, and I would certainly challenge it in the event it was ever used against me. Who wrote the program, and where did they get their data? (Rhetorical). As I stated in an earlier post, all impact info colllected from initiial crash testing has been proved worthless in the real world.

Rolebama, I understand your concerns, but the programmers have been sensible and honest in that respect; the results are presented as probabilities. This, even if you filmed the crash, is the best that can be said, "This is what probably happened". Contrary to what tends to happen at the moment, even with professional crash investigators reports, they are usually interpreted as black or white. Part of this may well be that normally what is being looked at is, can we charge any drivers with any of a list of motoring offences, rather than can we understand what happened.
As to the software, as mentioned before it uses data (from the manufacturers and test data) on how each car deforms and how much force is required, the laws of physics, high accuracy GPS data, photographs of the scene, weather conditions, and so on.
I saw one being used on TV, the cop using it said he liked it because the software went methodically through all the input data required quickly not allowing him to forget anything. As he pointed out, forgetting frequently happens, as they have to clear up the scene as quickly as possible.

As an interesting demonstration of how well the world can be modelled obeying the laws of physics try down loading a demonstration copy of The World of Goo from this site. http://2dboy.com/games.php

[Snowball]

In my view, all that any such software can achieve are things like contact speed, angle of contact, probable speed of vehicle(s) immediately prior to crash.
In no way can it accurately assess the circumstances, and who did what, before the activities which committed the vehicles to be certain to crash. Often, the prime mover in an RTC can themselves end up as not being part of the collision, and be left the scene before involved parties start to sort out what really happened.
The best evidence is reliable witnesses, which occur only in the luckiest circumstances.
Next comes the statements of those involved. This is where such software may be of use, since false statements may show that these cannot be true, based upon the post-crash condition of the vehicle(s).

The key role of insurance personnel is to define degree of blame, and settle accordingly.
If a prosecutable offence has been committed, this is a matter for the police/courts; the only interests that the insurers will have in this is how it affects the degree of blame (cost exercise between insurers), and whether to load next premium or even refuse cover.

In the final analysis, the insurance companies will operate under procedures which are satisfactory to themselves, and most economic to the business. If they do make use of any other software facilities, these will be used as and when they suit the insurers' findings; not as a blanket procedure for all cases.