Relation of having an objective

[Mttbnchtt]

Is there a relation to merely say that a planned process has an objective? "achieves planned objective" and "objective achieved by" will not work, because I do not want to also say that the planned process will succeed.

[bpeters42]

For every planned process, there is an implied 'plan specification' that 'plan specification' is in a 'has_part' relationship to an 'objective specification'. If you want to talk about the process before knowing if it is successful or not, the best was is to talk about the plan specification instead. Rather than duplicating a hierarchy of plan specifications and planned processes, we focus on the later though; if this is not clear, maybe a concrete example of what you had in mind will help.

[Mttbnchtt]

Thanks very much @bpeters42. Suppose I have a process P=planned process to improve diet and an objective O=improve diet. P is a class and O is an instance of an objective specification. How do you suggest to indicate that O is the objective of P?

[bpeters42]

I have to make some assumptions in writing this out; like 'improved compared to what', and if this is physician or trainer executing the plan, or if it is a self-improvement kind of goal. So let's take as a plan specification: "Improving my diet by eating less burgers and more vegetables". As written down, this is an instance of a plan specification, a type of information content entity (ICE). the part that says 'Improving my diet', is an instance of an objective specification. But you could also write "My goal is to eat better food by no longer shopping at Walmart". That is an instance of a different ICE. And the objective is written out in different word. But the intent is the same. So these would be different instantiations of the class "improving diet objective".

After I personally commit to " Improving my diet by eating less burgers and more vegetables", and not just write it down or tell people about it, it becomes a 'plan' in my head that I intend to execute. When I execute it successfully, that an instance of the planned process. That is where you would use the 'achieves planned objective' relationship.

We don't deal in our modeling with 'stupid plans' - e.g. "Improving my diet by only eating nails". We also don't deal with attempts that fail. Or at least: We prefer to describe attempts that succeed, and then define failures by how they differ from success. This is the same with e.g. anatomy, where ontologies define a canonical individual who has e.g. 2 kidneys, which still enables to talk about someone 'missing' a kidney, and infer that they have 1 and not 10 or 0).

Starting to ramble, and not sure how helpful this is. Would be happy to follow up.

On Tue, Jun 11, 2019 at 5:36 PM Matteo notifications@github.com wrote:

Thanks very much @bpeters42 https://github.com/bpeters42. Suppose I have a process P=planned process to improve diet and an objective O=improve diet. P is a class and O is an instance of an objective specification. How do you suggest to indicate that O is the objective of P?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/obi-ontology/obi/issues/1033?email_source=notifications&email_token=ADJX2IUSD7O4UY3PJE4SYRDP2BAKBA5CNFSM4HXC3Y4KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODXO4ZFI#issuecomment-501075093, or mute the thread https://github.com/notifications/unsubscribe-auth/ADJX2IRBAD5ZOY6CKZTQZFDP2BAKBANCNFSM4HXC3Y4A .

-- Bjoern Peters Professor La Jolla Institute for Allergy and Immunology 9420 Athena Circle La Jolla, CA 92037, USA Tel: 858/752-6914 Fax: 858/752-6987 http://www.liai.org/pages/faculty-peters

[Mttbnchtt]

Thanks so much @bpeters42. Let me call @pbuttigieg and @marieALaporte , who works with me, in this conversation. You wrote: "When I execute it successfully, that [becomes] an instance of the planned process". Can I have a planned process that is still under execution and it is unclear whether it will succeed or not?

[bpeters42]

I see no problem with it, but we have not spend a lot of time working out the details. It would still be a process, you can still talk about the plan that whoever is acting is pursuing. I guess I would create an 'incomplete planned process', which would be a sibling (not a child) of 'planned process'. I would still recommend for you to consider modeling the completed planned processes you have in mind, and then relate your ongoing processes to the parts of the completed one. Hope this helps a bit.

On Thu, Jun 13, 2019 at 2:38 PM Matteo notifications@github.com wrote:

Thanks so much @bpeters42 https://github.com/bpeters42. Let me call @pbuttigieg https://github.com/pbuttigieg and @marieALaporte https://github.com/marieALaporte , who works with me, in this conversation. You wrote: "When I execute it successfully, that [becomes] an instance of the planned process". Can I have a planned process that is still under execution and it is unclear whether it will succeed or not?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/obi-ontology/obi/issues/1033?email_source=notifications&email_token=ADJX2ITM3ELQINPWUCW6M53P2K46FA5CNFSM4HXC3Y4KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODXVDNQQ#issuecomment-501888706, or mute the thread https://github.com/notifications/unsubscribe-auth/ADJX2IQ7ML4WTU55FPRPB7LP2K46FANCNFSM4HXC3Y4A .

-- Bjoern Peters Professor La Jolla Institute for Allergy and Immunology 9420 Athena Circle La Jolla, CA 92037, USA Tel: 858/752-6914 Fax: 858/752-6987 http://www.liai.org/pages/faculty-peters

[pbuttigieg]

Thanks @bpeters42

We're dealing with multi-year development processes that of course have objectives and are planned, but are not complete or likely to be complete before we need an ontology class for them.

Looking at editor notes like this:

We are only considering successfully completed planned processes. A plan may be modified, and details added during execution. For a given planned process, the associated realized plan specification is the one encompassing all changes made during execution. This means that all processes in which an agent acts towards achieving some objectives is a planned process.

...does suggest that the process need not be complete.

The curator note:

This class merges the previously separated objective driven process and planned process, as they the separation proved hard to maintain. (1/22/09, branch call)

...is informative. "objective driven process" would be the right one in this case, I think.

For the moment, we're going to use 'planned process' for ongoing processes. If you say that this is incorrect, I think we should add this as an issue to an Operations Committee call.

[bpeters42]

By saying: "I would still recommend for you to consider modeling the completed planned processes you have in mind, and then relate your ongoing processes to the parts of the completed one" I meant to suggest exactly what you are doing: Use the planned processes as they are, that is perfectly fine. It is just that you will have axioms that imply things such as 'at the end of this investigation, we will have gathered data on XYZ'. If it ends up that the investigation gets cancelled before that data is gathered, you could reclassify the instance as a 'failed investigation'. But while a planned process is ongoing it is normally unproblematic to refer to it as a planned process. If you have specific problems that arise from that, I would be very curious to hear about them.

Maybe stupid analogy: If you are seeing a hand sticking out from a blanket, would you have a problem describing it using the FMA term for hand? FMA assumes that the hand is part of a body. But it could be that you are looking at a hand that was chopped off. If you are working on an ontology of bodily mutilation, these might be important distinctions, and I would question your life choices. But in most applications, it is perfectly fine and useful to refer to anatomical entities as parts of a canonically defined whole. We have used the same approach for planned processes.

• Bjoern

On Thu, Jul 11, 2019 at 12:04 PM Pier Luigi Buttigieg < notifications@github.com> wrote:

Thanks @bpeters42 https://github.com/bpeters42

We're dealing with multi-year development processes that of course have objectives and are planned, but are not complete or likely to be complete before we need an ontology class for them.

Looking at editor notes like this:

We are only considering successfully completed planned processes. A plan may be modified, and details added during execution. For a given planned process, the associated realized plan specification is the one encompassing all changes made during execution. This means that all processes in which an agent acts towards achieving some objectives is a planned process.

...does suggest that the process need not be complete.

The curator note:

This class merges the previously separated objective driven process and planned process, as they the separation proved hard to maintain. (1/22/09, branch call)

...is informative. "objective driven process" would be the right one in this case, I think.

For the moment, we're going to use 'planned process' for ongoing processes. If you say that this is incorrect, I think we should add this as an issue to an Operations Committee call.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/obi-ontology/obi/issues/1033?email_source=notifications&email_token=ADJX2IUBQF5WZLDEJQTOH63P6574FA5CNFSM4HXC3Y4KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODZXVVQY#issuecomment-510614211, or mute the thread https://github.com/notifications/unsubscribe-auth/ADJX2ITCO3IMVNEO3JALI5TP6574FANCNFSM4HXC3Y4A .

-- Bjoern Peters Professor La Jolla Institute for Allergy and Immunology 9420 Athena Circle La Jolla, CA 92037, USA Tel: 858/752-6914 Fax: 858/752-6987 http://www.liai.org/pages/faculty-peters

[pbuttigieg]

Thanks @bpeters42

I get the analogy, but still note that the lack of a simple, "objective-driven process" class is forcing several assumptions that seem unneeded in the SDGIO case (and others too, I suspect).

I suppose this wasn't felt before as the main focus was investigations that can be closed/classed as failed, rather than mulit-decade global development processes that tend to have much more involved evaluations/outcomes.

This interpretation of planned process:

.... all processes in which an agent acts towards achieving some objectives is a planned process.

Is the most workable I think.

Just because the process failed doesn't mean it didn't go to (or at least follow a) plan, and I don't see why 'failed processes' should be reclassified when 'successful processes' need not be. I would think being more neutral here (i.e. the planned process just has to follow a plan, it's success or failure are secondary) can lead to more consistency.

It is just that you will have axioms that imply things such as 'at the end of this investigation, we will have gathered data on XYZ'

I see the issue, stemming from the assumption that the planned process "worked" and delivered the expected outputs. Having axioms like "has intended output" on the class level would be better here. The instance-level axioms can then be used to autoclassify success or failure.

[bpeters42]

Very quickly:

• 'failed according to plan' id definitely a planned process. Say you have a drug trial, and it doesn't show efficacy, that is definitely still a process in which the plan succeeded. Sometimes failures are not considered when writing out a plan, but are implied. Say: You are planning a study to go to do observations in the antarctic in a specific year, and then there is a government shutdown that stops all flights. And you do it next year instead. You can add retrospectively to the study design that 'if idiots shut down the government in a given year, we will continue observations in the year after'. All of these are planned processes, and even if completed differently than originally anticipated, that doesn't really cause problems. However, if you have an investigation in year 3, where you find out that key data being generated as part of the study is being made-up by a rogue scientist, it becomes really hard to still consider this part of the plan. That is what we would consider a failed investigation. Same with a meteor destroying the sample repository that you have been generating etc. If you are abandoning a process that was originally objective driven for an abnormal reason, that is what we would consider a 'failed process'.

On Fri, Jul 12, 2019 at 7:32 AM Pier Luigi Buttigieg < notifications@github.com> wrote:

Thanks @bpeters42 https://github.com/bpeters42

I get the analogy, but still note that the lack of a simple, "objective-driven process" class is forcing several assumptions that seem unneeded in the SDGIO case (and others too, I suspect).

I suppose this wasn't felt before as the main focus was investigations that can be closed/classed as failed, rather than mulit-decade global development processes that tend to have much more involved evaluations/outcomes.

This interpretation of planned process:

.... all processes in which an agent acts towards achieving some objectives is a planned process.

Is the most workable I think.

Just because the process failed doesn't mean it didn't go to plan, and I don't see why 'failed processes' should be reclassified when 'successful processes' need not be. I would think being more neutral here (i.e. the planned process just has to follow a plan, it's success or failure are secondary) can lead to more consistency.

It is just that you will have axioms that imply things such as 'at the end of this investigation, we will have gathered data on XYZ'

I see the issue, stemming from the assumption that the planned process "worked" and delivered the expected outputs. Having axioms like "has intended output" on the class level would be better here. The instance-level axioms can then be used to autoclassify success or failure.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/obi-ontology/obi/issues/1033?email_source=notifications&email_token=ADJX2IXVN56AFA4UJUP4FR3P7CIXTA5CNFSM4HXC3Y4KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODZZ5UNQ#issuecomment-510908982, or mute the thread https://github.com/notifications/unsubscribe-auth/ADJX2IVLIP2TA3MYO7C6H3TP7CIXTANCNFSM4HXC3Y4A .

-- Bjoern Peters Professor La Jolla Institute for Allergy and Immunology 9420 Athena Circle La Jolla, CA 92037, USA Tel: 858/752-6914 Fax: 858/752-6987 http://www.liai.org/pages/faculty-peters

[pbuttigieg]

@bpeters42 @cmungall just an update - in SDGIO (and I'm sure ontologies in other domains) we need a simpler and more useful relation and handling of plans, objectives, and action specifications.

All we need so say is that open-ended processes are following a plan. That's it. (I say this realising we may be hitting the limits of expression with OWL)

I guess I would create an 'incomplete planned process', which would be a sibling (not a child) of 'planned process'.

We're aiming at a more general level where there doesn't need to be an assertion of completeness/incompleteness. By that logic, you'd have:

• planned process
  • completed planned process
    • successfully completed planned process
    • unsuccessfully completed planned process
  • incomplete planned process

The 2009 editor notes on directive information entity show that there was an attempt to make things more usable, but I don't see any such outcomes.

What we will do:

• We're creating a placeholder relation in SDGIO for "has objective specification" which will have domain 'planned process' and range 'objective specification', this will allow to query efficiently.
• We will not create another planned process class (yet) to avoid more turbulence for the moment, but reiterate that something must to done to make this usable, hopefully in OBO Core.
• We are treating the existing planned processes in the most general and natural sense - processes that follow some plan.

The notes on the classes here suggest that there should be an actual class under realizable entity for things like plans, objectives, etc. with relations like "has objective" mirroring semantics of "has role" or "has dispostion".

[cmungall]

I would shy away from baking in an orthogonal axis of classification under process. Have a separate set of properties that can apply to processes (can also be modeled as parts of processes, e.g. a successful completion, a failure point).

Reason: Separation of concerns. Different use cases may require different ways of slicing and dicing this, makes sense to do this in an orthogonal hierarchy.

[pbuttigieg]

@cmungall recommendations noted.

Having "successful completion" is the same as being a subclass of "successfully completed planned process", with a similar set up for failure points and being a failed process. So I don't really see the difference or use of orthogonal structures here.

Reason: Separation of concerns. Different use cases may require different ways of slicing and dicing this, makes sense to do this in an orthogonal hierarchy.

I can't think of any use case that has a plan guiding a process and is not concerned with completion and success/failure. I think this is stable enough to be baked in.

If for some reason there is no dimension of success or failure in the plan's objective spec, then it can be an instance of the superclass.

[linikujp]

Hi all,

Good discussion here!

We have example of a clinical trial, which is a planned process. There will be completed clinical trial, paused clinical trials, failed clinical trials, and currently ongoing (uncompleted) clinical trials. I think they are more like status of the executed plan at a certain time point.

We are struggling with linking the process (clinical trial) with part of the plan specification or objectives, such as an intervention specified in the plan specification (protocol). Any suggestions?

Thanks, Asiyah

On Fri, Aug 23, 2019 at 11:24 AM Pier Luigi Buttigieg < notifications@github.com> wrote:

@cmungall https://github.com/cmungall recommendations noted.

Having "successful completion" is the same as being a subclass of "successfully completed planned process", with a similar set up for failure points and being a failed process. So I don't really see the difference or use of orthogonal structures here.

Reason: Separation of concerns. Different use cases may require different ways of slicing and dicing this, makes sense to do this in an orthogonal hierarchy.

I can't think of any use case that has a plan guiding a process and is not concerned with completion and success/failure. I think this is stable enough to be baked in.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/obi-ontology/obi/issues/1033?email_source=notifications&email_token=ACCPEPIOJYLP2JI45ZPFFB3QF76MRA5CNFSM4HXC3Y4KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD5ARABQ#issuecomment-524357638, or mute the thread https://github.com/notifications/unsubscribe-auth/ACCPEPKQCRP5YHVTO4LTQI3QF76MRANCNFSM4HXC3Y4A .

[pbuttigieg]

Hi all,

This was discussed in today's OBO Operations call. Based on this issue, @bpeters42 @jamesaoverton sketched out a proposal to include ongoing, completed (successful), and failed planned processes. A softer relation between an objective specification and an ongoing planned process was also proposed.

There will be a few more steps in ironing this out, but this may help @Mttbnchtt needs as well as those raised by @linikujp

[jamesaoverton]

Sorry for the delay with this. I'm going to post a prototype to the OBO Core repo, then link to it here.

[cmungall]

I saw James' presentation on the OBO call and I am fully behind it (except some very minor points about what the string "plan" should denote)

[jamesaoverton]

This is the PR: https://github.com/OBOFoundry/Experimental-OBO-Core/pull/48

[bpeters42]

This is now in COB and discussions should continue there. Closing for OBI.