Page 1 of 1

Definitions

Posted: Mon Apr 29, 2024 6:45 pm
by Horse
Probably of little interest to anyone other than Spin and me :)

A discussion paper published recently, proposed interesting (re-) definitions of some terms (context is AVs) , from usual, such as:

- Safety: absence of unreasonable risk
• Unreasonable risk: risk judged to be unacceptable in a certain context according to valid societal moral concepts
• Risk: combination of the probability of occurrence of harm and the severity of that harm
• Severity: estimate of the extent of harm to one or more individuals that can occur in a potentially hazardous event
• Hazardous event: combination of a hazard and an operational situation
• Hazard: potential source of harm caused by malfunctioning behavior of the item
• Harm: physical injury or damage of persons
• Malfunctioning behavior: failure or unintended behavior of an item with respect to its design intent
• Operational situation: scenario that can occur


To:

- Acceptable: meets all safety constraints as shown by a safety case Note: The phrase “acceptably safe” might be used in some contexts. While “Safety” is used as a modifier, use of the word “safe” alone should be avoided.
• Safety case: structured argument, supported by a body of evidence, that provides a compelling, comprehensible, and sound argument that safety engineering efforts have ensured a system meets a comprehensive set of safety constraints. Note: This emphasizes meeting constraints rather than net risk. A limit to defined operational environments is intentionally excluded.
• Safety engineering: a methodical process of ensuring a system meets all its safety constraints throughout its lifecycle, including hazard analysis, risk assessment, risk mitigation, validation, and field engineering feedback. Note: Requires safety engineering beyond brute force test validation. Hazard analysis is broadened to address all safety constraints.
• Safety constraint: a limitation imposed on risk or other aspects of the system by stakeholder requirements . Note: This implicitly requires identification of stakeholders who might be affected by losses, and makes it more straightforward to view safety as a multi-dimensional constrained optimization problem rather than a mostly one-dimensional pure risk optimization problem [Koopman24b]. Safety constraints might include: limits on individual risks, limits on net risk, exposure limits for specified types of risk patterns, and issues that are difficult to trace to pure risk.
• Risk: combination of the probability of occurrence of a loss, or pattern of losses, and the importance to stakeholders of the associated consequences. Note: Consequence (severity) might be an overriding concern regardless of probability. Net importance can be non-linearly related to individual losses if
forming a pattern. Correlated loss events, inequitable loss patterns, and loss patterns involving a failure to mitigate emergent loss trends are in-scope.
• Loss: an adverse outcome, including damage to the system itself, negative societal externalities, damage to property, damage to the environment, injury or death to animals, and injury or death to people
Note: This is broader in scope than some other typical definitions of loss or harm. Some types of loss might be assigned very low severity in some application domains. Allocation of blame does not affect whether a loss occurred.

Re: Definitions

Posted: Mon Apr 29, 2024 9:04 pm
by Taff
• Safety engineering: a methodical process of ensuring a system meets all its safety constraints throughout its lifecycle, including hazard analysis, risk assessment, risk mitigation, validation, and field engineering feedback. Note: Requires safety engineering beyond brute force test validation. Hazard analysis is broadened to address all safety constraints.



That's my job description.

Re: Definitions

Posted: Mon Apr 29, 2024 9:18 pm
by Cousin Jack
Acceptably safe: doesn't worry me unduly,

Re: Definitions

Posted: Mon Apr 29, 2024 9:20 pm
by Cousin Jack
Hazard: dangerous to idiots.

Re: Definitions

Posted: Mon Apr 29, 2024 9:26 pm
by Horse
Cousin Jack wrote: Mon Apr 29, 2024 9:20 pm Hazard: dangerous to idiots.
Your definition of 'idiots'? :eh:

Re: Definitions

Posted: Mon Apr 29, 2024 10:18 pm
by Horse
Taff wrote: Mon Apr 29, 2024 9:04 pm
• Safety engineering: a methodical process of ensuring a system meets all its safety constraints throughout its lifecycle, including hazard analysis, risk assessment, risk mitigation, validation, and field engineering feedback. Note: Requires safety engineering beyond brute force test validation. Hazard analysis is broadened to address all safety constraints.
That's my job description.
I used to call myself a professional pessimist :)

Re: Definitions

Posted: Tue Apr 30, 2024 8:30 am
by Cousin Jack
Horse wrote: Mon Apr 29, 2024 9:26 pm
Cousin Jack wrote: Mon Apr 29, 2024 9:20 pm Hazard: dangerous to idiots.
Your definition of 'idiots'? :eh:
Obvious, innit.

Other people.
I do a finely judged overtake.
You are pushing your luck.
They are idiots with a death wish.

Simples

Re: Definitions

Posted: Tue Apr 30, 2024 10:38 am
by iansoady
Cousin Jack wrote: Mon Apr 29, 2024 9:18 pm Acceptably safe: doesn't worry me unduly,
It depends who is accepting it. If I was a potential victim my criteria for acceptability may differ from those of a potential perpetrator.

Mind you I don't have the energy to read through the whole set of definitions......

Re: Definitions

Posted: Tue Apr 30, 2024 10:44 am
by Horse
Ah, ok. I was taking it to its limits.

Something like a radiation leak is invisible to everyone.


Your definition sounds like the assessment and marking scheme on one of my training courses ;)

Re: Definitions

Posted: Wed May 01, 2024 5:54 pm
by The Spin Doctor
I'm reminded of Colin Chapman... his principle of 'simplify, then add lightness' applies to definitions too.

Re: Definitions

Posted: Wed May 01, 2024 8:08 pm
by Horse
The Spin Doctor wrote: Wed May 01, 2024 5:54 pm I'm reminded of Colin Chapman... his principle of 'simplify, then add lightness' applies to definitions too.
Go on then :D

Re: Definitions

Posted: Wed May 01, 2024 8:28 pm
by Taff
I sometimes feel that some people get too hung up with definitions and splitting hairs between a couple of words.
What really matters is that someone is actually doing a good job of the safety assessment process and developing good safety requirements.
In aerospace ARP4761 tells us exactly how to do this work and I regularly tell our design engineers our first goal is to do good engineering to get a good qualitative argument, and then we'll work out the quantitative argument (fault trees)

Re: Definitions

Posted: Thu May 02, 2024 11:52 pm
by The Spin Doctor
Horse wrote: Wed May 01, 2024 8:08 pm
The Spin Doctor wrote: Wed May 01, 2024 5:54 pm I'm reminded of Colin Chapman... his principle of 'simplify, then add lightness' applies to definitions too.
Go on then :D
I thought your first list was good.

Re: Definitions

Posted: Fri May 03, 2024 5:28 pm
by Horse
The Spin Doctor wrote: Thu May 02, 2024 11:52 pm
Horse wrote: Wed May 01, 2024 8:08 pm
The Spin Doctor wrote: Wed May 01, 2024 5:54 pm I'm reminded of Colin Chapman... his principle of 'simplify, then add lightness' applies to definitions too.
Go on then :D
I thought your first list was good.
To be honest, so did I :D

But presumably it's not sufficient for AVs.

Re: Definitions

Posted: Fri May 03, 2024 8:10 pm
by Taff
If you want to geek out on the real geekery of systems safety, have a mooch about in the publications on this website - the Safety Critical Systems Club. it's free to create a login.

there's loads and loads of papers dedicated to Autonomous Vehicles and their safety, one of the more recent ones talks around how we can determine what is an acceptable level of safety for an AV. It's a minefield.

https://scsc.uk/

Re: Definitions

Posted: Fri May 03, 2024 8:20 pm
by Horse
Taff wrote: Fri May 03, 2024 8:10 pm It's a minefield.
That's one of the reasons that MoD are developing AVs :thumbup: