Southern DOO: ASLEF members vote 79.1% for revised deal

[hounddog]

Human error, none of us are infallible

As the drivers and guards on here are so quick to remind everyone whenever one of their number gets something wrong.

 

[Tetchytyke]

Really? where did you see this?

The jungle telegraph tells us that it was a manager (acting as a guard) who gave two dings against a red signal. GTR are obviously being very coy about it.

I thought they were always open to talks in the first place?

Once again, I refer you to the recent letter from Charles Horton who says "GTR will not change their plans" and that "the RMT's demands cannot be met".

https://www.facebook.com/Confession...1126329037932/611125639038001/?type=3&theater

That doesn't sound very open to talks to me. That sounds like fighting talk to me.

 

[OliverS]

That is one of the biggest fallacies through the misuse of statistics.

You are talking about the probability of failure by Person 1 (say P1) or failure by Person 2 (say P2) resulting in a failure, which has no relevance to this incident at Chichester. What you need to be looking at is P1 and P2 resulting in a failure in service. (The probability of TPWS failure is irrelevant in this case, as it does not depend on either Person 1 or Person 2’s behaviour, and works independently whatever that probability is, assuming that it has not been tampered with.

So in the simplest form, you need to be looking at P1 (which would be less than 1, ie. always failing) multiplied by P2 (which is also less than 1), therefore the simplest form of mathematical theory would no doubt show that the probability of failure with two people present would be unquestionably lower than with only one person present.

What complicates matters here is that this simplest form assumes that the behaviour of Person 1 and Person 2 are independent, ie. the probability of failure by Person 1 is not affected by the behaviour of Person 2, and vice versa. Of course no matter how much we train ourselves and maintain our discipline, this is rarely the case with human behaviour, so in this case what we need to do is look at the probability of failure by Person 2 conditional on the event of failure by Person 1 (call it Q), and multiply P1 by this conditional probability Q. We only need to consider this case as with train dispatch, the actions are sequential, ie. Person 1 (guard) giving two on the buzzer before Person 2 (driver) performing final checks and applying power.

There are a lot of theories out there which argue that Q is much closer to 1 than P2 is, which is quite understandable. What is not clear is the relationship between P1 x Q (ie. two-person dispatch) and P2 (ie. driver only dispatch), which is part of the reason why this topic always attracts lots of discussion and no conclusive statistical proof either way.

In any case this is still a simplified model, but forms pretty much the fundamental structure of modelling such risks. The difficulty for any practitioner is to come up with accurate figures for these measures.

This does not mean some of your other arguments are wrong, just that I cannot let this go unchallenged as it is one of the biggest mistakes people make in the application of probability, and arguments based on this logic would be null and void. In fact, I quite agree that this isolated incident does not show the inherent danger of anything by itself, although I think most people can agree that someone with only 5 days' training is likely to be less experienced than someone with 6 months' training. I am unable to say whether that makes them more prone to mistakes, although the timing of this incident is really rather unfortunate.

I'm not sure I entirely agree with you, although you are applying the general statistical rule more tightly to dispatch than I had intended. I was originally disagreeing with the notion that adding a second person halved the risk. It does not as adding a second person increases the number of failure modes. You are considering only one failure mode not any failure of the system where a failure is anything that causes the train not to run.

Briefly, adding complexity adds failure modes. KISS is normally the best and safest way.

 

[Llama]

If individuals have been trained to meet the Rule Book requirements of the Guard, there is diddly the Driver could do. In the eyes of the safety governance that goes across the whole railway, the individual is competent.

...until they prove themselves not competent, by performing the cardinal sin of giving RTS against a red. If such an action is reported, the usual procedure is then to remove the member of staff from safety critical duties and then have them med-screened and interviewed.

 

[Carlisle]

GTR want to talk about the OBS role, RMT want to to talk about keeping guards.

So are you able to suggest a possible compromise that doesn't involve either side pretty much forcing the other into total submission?

 

[bb21]

Briefly, adding complexity adds failure modes. KISS is normally the best and safest way.

In that case you are adding another factor into play, but using the wrong statistical concept to justify it.

I think I know now what you were trying to say, but there is no easy way of justifying it using a dumbed down statistical model or any general rule. "Adding another person into the mix may increase risks of overall failure" is probably a better way of summarising it.

 

[Agent_c]

I'm not sure I entirely agree with you, although you are applying the general statistical rule more tightly to dispatch than I had intended. I was originally disagreeing with the notion that adding a second person halved the risk. It does not as adding a second person increases the number of failure modes. You are considering only one failure mode not any failure of the system where a failure is anything that causes the train not to run.

Briefly, adding complexity adds failure modes. KISS is normally the best and safest way.

So you ARE advocating for an increased risk of catastrophic failure because of some potential minor inconvenience of a non-stood-down staffer being available.

Exactly how many minutes delay is a human life worth these days?
--- old post above --- --- new post below ---

So are you able to suggest a possible compromise that doesn't involve either side pretty much forcing the other into total submission?

I would suggest driver ordinarily controls doors allowing OBS to sell more tickets, OBS remains safety trained to allow for evacuation and can dispatch if the driver feels circumstances warrant it.

 

[OliverS]

In that case you are adding another factor into play, but using the wrong statistical concept to justify it.

I think I know now what you were trying to say, but there is no easy way of justifying it using a dumbed down statistical model or any general rule.

I was disagreeing with post 944 that said:

I think your logic is faulty.
If two people have to agree to "go" and both have to check the signal, then the chance of failure has halved.

The point being that when only one failed, i.e. the guard, then the train failed from a whole system point of view. Failure is not just restricted to SPADs.

Adding the guard, as you pointed out, does not significantly seem to affect dispatch failures. Hence it does not half the chance of failure.

 

[bb21]

The point being that when only one failed, i.e. the guard, then the train failed from a whole system point of view. Failure is not just restricted to SPADs.

Adding the guard, as you pointed out, does not significantly seem to affect dispatch failures. Hence it does not half the chance of failure.

No, it probably does not half the chance of failure for a number of reasons. What obviously is also quite important is whether it reduces risks (even if not by half), and how that weighs against the other roles guards play and the cost implications.

I don't profess to know the exact answers and can only form an opinion on a number of other data sources (many only qualitative). Perhaps GTR could enlighten us.

Thinking about your definition of "failure" a bit more, there are loads more factors that could impact on it, so the variation of the measure in different contexts could be quite staggering. Interesting.

 

[Agent_c]

I was disagreeing with post 944 that said:



The point being that when only one failed, i.e. the guard, then the train failed from a whole system point of view. Failure is not just restricted to SPADs.

Adding the guard, as you pointed out, does not significantly seem to affect dispatch failures. Hence it does not half the chance of failure.

That's because you're using faulty logic that one failure is as bad as another. It's not. Having two pilots in the cockpit increases the chance the flight will need to be abandoned if one falls ill - but if someone is going to fall ill do you want a second man there or not?

Answer before you crash please.

 

[Bletchleyite]

That's because you're using faulty logic that one failure is as bad as another. It's not. Having two pilots in the cockpit increases the chance the flight will need to be abandoned if one falls ill - but if someone is going to fall ill do you want a second man there or not?

Answer before you crash please.

Trains are different from planes, because if the driver loses consciousness they stop, not crash.

A train stopped between stations for a few hours while someone gets to it (which does happen with guards if an evacuation is not easy to organise or if it is felt incorrectly that things might move soon) is, by and large, an inconvenience, not a mass of tangled metal and bodies on the floor.

 

[Domh245]

There is a reason that in safety critical maintenance on aircraft, someone does the work and then it is checked by another person before being ok-ed. In a situation where you have 2 people each doing half of a task, then yes it increases the risk of something going wrong, but if it is 2 people checking something, it is less likely to go wrong as someone else is there to pick up on errors. Its similar to how things will have 2 independent sensors to check things like speed or temperatures, so that if 1 makes a mistake, the other can spot it and dispute it.

 

[Bletchleyite]

There is a reason that in safety critical maintenance on aircraft, someone does the work and then it is checked by another person before being ok-ed. In a situation where you have 2 people each doing half of a task, then yes it increases the risk of something going wrong, but if it is 2 people checking something, it is less likely to go wrong as someone else is there to pick up on errors. Its similar to how things will have 2 independent sensors to check things like speed or temperatures, so that if 1 makes a mistake, the other can spot it and dispute it.

Indeed so. As a Scout climbing instructor who sometimes instructs alone (with another non skilled adult from the group there) and sometimes in a pair, I certainly feel safer with a second instructor as we can double-check each others' actions.

 

[Agent_c]

Trains are different from planes, because if the driver loses consciousness they stop, not crash.

If it was a DOO driver, not a management guard who didn't see the red and SPADed, he potentially could have crashed.

A train stopped between stations for a few hours while someone gets to it (which does happen with guards if an evacuation is not easy to organise or if it is felt incorrectly that things might move soon) is, by and large, an inconvenience, not a mass of tangled metal and bodies on the floor.

You're saying that a train crash cannot end up in a tangled mess of metal and bodies? Are you *sure* about that?

 

[Bletchleyite]

You're saying that a train crash cannot end up in a tangled mess of metal and bodies? Are you *sure* about that?

No, what I'm pointing out is that with a train, if in doubt you can stop, near enough anywhere you like, and this can be carried out automatically. If this occurs for some time it is nothing other than an inconvenience. You cannot stop an aircraft unless you have somewhere to land it. There has to be a functioning pilot, or it will crash with huge loss of life.

 

[Pumbaa]

...until they prove themselves not competent, by performing the cardinal sin of giving RTS against a red. If such an action is reported, the usual procedure is then to remove the member of staff from safety critical duties and then have them med-screened and interviewed.

Rule Book Competence shouldn't be confused with competence at a job. Any Guard who gives the ready to start signal against a red is still competent (with regard to the Rule Book) but may have just demonstrated that they are an incompetent to someone.

 

[Tetchytyke]

Briefly, adding complexity adds failure modes.

But there are different types of failure. The only ones that really truly count are the ones where a danger becomes present.

Talking about despatch, the relevant failures are a driver taking power against a red signal or a passenger becoming trapped in the doors. Everything else is just an inconvenience.

I'd agree that having two people doing despatch does not "halve" the risk.

However, as bb21 has explained, it does reduce the risk. If 0=no probability and 1=guaranteed probability, then the risk of the guard dinging against the red is x<1, and the risk of the driver taking power against the red is y<1. The combined probability is therefore x*y, which must be lower than either x or y in isolation.

 

[Wolfie]

Consider, if there is one person and their chance of failure is P then that is also the chance of failure. If there are two people both with a chance of failure of P then the total chance is the chance that the first person will fail added to the chance that they won't fail but the second person will. I.e. adding the second person makes failure more likely not less.

Sorry but with the greatest of respect this is utter and complete rubbish!

That is one of the biggest fallacies through the misuse of statistics.

You are talking about the probability of failure by Person 1 (say P1) or failure by Person 2 (say P2) resulting in a failure, which has no relevance to this incident at Chichester. What you need to be looking at is P1 and P2 resulting in a failure in service. (The probability of TPWS failure is irrelevant in this case, as it does not depend on either Person 1 or Person 2’s behaviour, and works independently whatever that probability is, assuming that it has not been tampered with.

So in the simplest form, you need to be looking at P1 (which would be less than 1, ie. always failing) multiplied by P2 (which is also less than 1), therefore the simplest form of mathematical theory would no doubt show that the probability of failure with two people present would be unquestionably lower than with only one person present.

But there are different types of failure. The only ones that really truly count are the ones where a danger becomes present.

Talking about despatch, the relevant failures are a driver taking power against a red signal or a passenger becoming trapped in the doors. Everything else is just an inconvenience.

I'd agree that having two people doing despatch does not "halve" the risk.

However, as bb21 has explained, it does reduce the risk. If 0=no probability and 1=guaranteed probability, then the risk of the guard dinging against the red is x<1, and the risk of the driver taking power against the red is y<1. The combined probability is therefore x*y, which must be lower than either x or y in isolation.

Yup - and I realise that my coin example might be taken as suggesting I think having two staff presnet might half the risk - it ain't so!

What complicates matters here is that this simplest form assumes that the behaviour of Person 1 and Person 2 are independent, ie. the probability of failure by Person 1 is not affected by the behaviour of Person 2, and vice versa. Of course no matter how much we train ourselves and maintain our discipline, this is rarely the case with human behaviour, so in this case what we need to do is look at the probability of failure by Person 2 conditional on the event of failure by Person 1 (call it Q), and multiply P1 by this conditional probability Q. We only need to consider this case as with train dispatch, the actions are sequential, ie. Person 1 (guard) giving two on the buzzer before Person 2 (driver) performing final checks and applying power.

There are a lot of theories out there which argue that Q is much closer to 1 than P2 is, which is quite understandable. What is not clear is the relationship between P1 x Q (ie. two-person dispatch) and P2 (ie. driver only dispatch), which is part of the reason why this topic always attracts lots of discussion and no conclusive statistical proof either way.

In any case this is still a simplified model, but forms pretty much the fundamental structure of modelling such risks. The difficulty for any practitioner is to come up with accurate figures for these measures.

This does not mean some of your other arguments are wrong, just that I cannot let this go unchallenged as it is one of the biggest mistakes people make in the application of probability, and arguments based on this logic would be null and void. In fact, I quite agree that this isolated incident does not show the inherent danger of anything by itself, although I think most people can agree that someone with only 5 days' training is likely to be less experienced than someone with 6 months' training. I am unable to say whether that makes them more prone to mistakes, although the timing of this incident is really rather unfortunate.

Thank the Lord that someone on here actually understand stats!

To give a trivial example: The probability, when tossing a coin, of getting a head is 1/2 (or 0.5). If tossing two coins the probability of getting 2 heads is 0.5 squared ie 0.25 or 1/4. The example is similar because each coin operates independently. In a smilar way each staff member does their job individually and independently.

But there are different types of failure. The only ones that really truly count are the ones where a danger becomes present.

Talking about despatch, the relevant failures are a driver taking power against a red signal or a passenger becoming trapped in the doors. Everything else is just an inconvenience.

I'd agree that having two people doing despatch does not "halve" the risk.

However, as bb21 has explained, it does reduce the risk. If 0=no probability and 1=guaranteed probability, then the risk of the guard dinging against the red is x<1, and the risk of the driver taking power against the red is y<1. The combined probability is therefore x*y, which must be lower than either x or y in isolation.

Yup and I realise that my coin example might inadvertently be taken as implying that having two staff members presnet halves the risk - it ain't so!

 

[Bletchleyite]

Rule Book Competence shouldn't be confused with competence at a job. Any Guard who gives the ready to start signal against a red is still competent (with regard to the Rule Book) but may have just demonstrated that they are an incompetent to someone.

The meaning of "Competent" (to a set of regulations) and "competent" (ability to actually do the job right all the time) often differ in many industries. There are, for instance, an awful lot of tradesmen of various kinds who may be legally Competent but certainly are not competent. Meanwhile, I am competent at home electrical work (I know how to do it correctly and safely) but am not Competent at it (I am not Part P registered, for instance).

 

[KM1991]

What I'm about to say is related in a way to DOO on this GTR franchise. According to station line managers, once the class 700 comes into service on the thameslink route, all dispatch duties will be taken away from station staff (where applicable) for these trains, but continue on the 319/377 class for the time being. I haven't really heard much from the RMT/Aslef about this, anyone else heard anything? A potential dispute maybe...

 

[hwl]

The jungle telegraph tells us that it was a manager (acting as a guard) who gave two dings against a red signal. GTR are obviously being very coy about it.



Once again, I refer you to the recent letter from Charles Horton who says "GTR will not change their plans" and that "the RMT's demands cannot be met".

https://www.facebook.com/Confession...1126329037932/611125639038001/?type=3&theater

That doesn't sound very open to talks to me. That sounds like fighting talk to me.

I got an email for GTR that said:
"... we are pleased that the RMT has decided to attend talks this Friday ... "

 

[Agent_c]

No, what I'm pointing out is that with a train, if in doubt you can stop, near enough anywhere you like, and this can be carried out automatically. If this occurs for some time it is nothing other than an inconvenience. You cannot stop an aircraft unless you have somewhere to land it. There has to be a functioning pilot, or it will crash with huge loss of life.

Could you please point me to a train that can stop at a moment notice, rather than having to deal with Newton's laws of physics?

 

[Bletchleyite]

Could you please point me to a train that can stop at a moment notice, rather than having to deal with Newton's laws of physics?

That wasn't what I meant, and in fact an automated system will not initiate a stop any faster or slower than a driver.

Meanwhile, planes do not land themselves in the event of the pilot(s) being incapacitated (though we can't be far off that).

 

[cjmillsnun]

I got an email for GTR that said:
"... we are pleased that the RMT has decided to attend talks this Friday ... "

Yes but what will happen is GTR will insist on their way or the highway and nothing will get resolved.

 

[Tetchytyke]

Could you please point me to a train that can stop at a moment notice, rather than having to deal with Newton's laws of physics?

Neil's point is that if a driver becomes incapacitated "at the wheel" the train's safety systems will apply emergency braking and stop. Given the way signalling works it will stop safely. If a pilot becomes incapacitated "at the wheel" then there is nobody to fly the thing and it will (eventually) crash; this is the hypothesis for what happened to MH17.

Thank the Lord that someone on here actually understand stats!

To give a trivial example: The probability, when tossing a coin, of getting a head is 1/2 (or 0.5). If tossing two coins the probability of getting 2 heads is 0.5 squared ie 0.25 or 1/4. The example is similar because each coin operates independently. In a smilar way each staff member does their job individually and independently.

However that's mathematically irrelevant. Whatever the probability is of x and whatever the probability is of y, x*y must be lower than the probability of x or y in isolation. Even where x=1 (as it would be if the guard dings against a red) x*y<1.

The complication here is that x and y are not truly independent, as the driver can be influenced by the behaviour of the guard. There have been several cases of ding-ding-and-away where the driver has responded to the guard without checking himself.

The difficulty therefore is proving x*y<z, where z is the probability of the DOO driver closing doors and taking power against a red. But one would assume, at the very least, that x=z, i.e. the chance of a driver and a guard making the same mistake are the same.

 

[Pumbaa]

The meaning of "Competent" (to a set of regulations) and "competent" (ability to actually do the job right all the time) often differ in many industries. There are, for instance, an awful lot of tradesmen of various kinds who may be legally Competent but certainly are not competent. Meanwhile, I am competent at home electrical work (I know how to do it correctly and safely) but am not Competent at it (I am not Part P registered, for instance).

Quite. My original point being that drivers couldn't refuse to work with management Guards because they were Competent people.

 

[Tetchytyke]

I got an email for GTR that said:
"... we are pleased that the RMT has decided to attend talks this Friday ... "

It'll be interesting, won't it, if Horton has the same attitude on Friday. Because, as it stands, he's basically saying "my way or the highway".

 

[Skimble19]

What I'm about to say is related in a way to DOO on this GTR franchise. According to station line managers, once the class 700 comes into service on the thameslink route, all dispatch duties will be taken away from station staff (where applicable) for these trains, but continue on the 319/377 class for the time being. I haven't really heard much from the RMT/Aslef about this, anyone else heard anything? A potential dispute maybe...

This is a very good question, and one which management on the GN side are being incredibly coy about...

 

[FordFocus]

What I'm about to say is related in a way to DOO on this GTR franchise. According to station line managers, once the class 700 comes into service on the thameslink route, all dispatch duties will be taken away from station staff (where applicable) for these trains, but continue on the 319/377 class for the time being. I haven't really heard much from the RMT/Aslef about this, anyone else heard anything? A potential dispute maybe...

Glad to see the Govia and Horton strategy of destaffing is reaching more grades. :roll:

The 700s and 377 class are similar in respects, they are fitted with DOO onboard cameras, so what makes the 700s more safer?

 

[Drogba11CFC]

Yes but what will happen is GTR will insist on their way or the highway and nothing will get resolved.

Leading to further industrial action, and potentially spreading to other TOCs as well.