Scientific knowledge dependency graph

[quitequick]

I'm asking this question here on a physics forum because a physics question sparked this off. However, it is applicable to all aspects of scientific knowledge.

I am looking for a scientific knowledge dependency graph. This is one centralised database which captures the specific and required dependencies between proven scientific hypotheses. For example, hypothesis Z is only proven if hypotheses W, X, and Y are proven. In turn, hypothesis W is only proven if hypotheses P, Q and R are proven. Etcetera. From one hypothesis I would hope to easily trace the chain/graph/hierarchy of required proven hypotheses all the way back to, perhaps, mathematical axioms and 'fundamental' observations.

Does such a thing exist?

 

[SteamKing]

I doubt it. As physics is always adding new knowledge, it often also requires the addition of new hypotheses or the modification of previously proposed hypotheses. Any such structure as you are searching for would be in a constant state of change.

Imagine if such a structure existed in say 1899. A few years later, after a few physics minds had been blown, an entirely new structure would have required construction as various quantum theories were proposed and the structure of the atom was investigated in more detail.

 

[harrylin]

[..] I am looking for a scientific knowledge dependency graph. This is one centralised database which captures the specific and required dependencies between proven scientific hypotheses. [..]

I doubt it. As physics is always adding new knowledge, it often also requires the addition of new hypotheses or the modification of previously proposed hypotheses. Any such structure as you are searching for would be in a constant state of change. [..]

Obviously; and that's the idea that I got from the question. Software dependencies are in a constant state of change with new programs and new operating system versions coming out all the time.
So I searched a little (wow this thread was already found!), and did not find much...

Probably what you are looking for is too complex (and too much effort) to be doable in practice (not a big market for it I'm afraid).

Here's one thing that I found:
http://www.redorbit.com/news/science/152611/mapping_scientific_knowledge_in_diagram_form/

It appears that it is in fact a program to be filled in (maybe some people did) and not a completed diagram. However, that points to the new search terms "cmaps" and "physics concept maps" - maybe there are some general physics concept maps around?

As a matter of fact, that last search term gives many interesting hits, with among them a diagram in a site that we know rather well at this forum (scroll down to the second diagram):
http://hyperphysics.phy-astr.gsu.edu/hbase/hph.html

Does that help?

 

[quitequick]

Does that help?

Thanks, Harry - yes it helps. But I suppose I am looking for more something more rigorous than links between concepts. I'm interested in verifiable hypothesis dependencies - so if someone publishes a paper that, say, derives from other's works, those other works also have to be proven, and so on. So, if a hypothesis subsequently proves incorrect (new evidence?) then all works that are dependent also become incorrect, or at least, subject to re-proof. An audit trail, if you will.

Probably what you are looking for is too complex (and too much effort) to be doable in practice (not a big market for it I'm afraid).

Yes its complex, it will be dynamic and may not have any immediate commercial value. But that's no reason not have one!

It sounds to me like the scientific process should require it. As our knowledge grows, which often time is built upon other knowledge, we need to map this growth to be sure that all new knowledge is always founded on verifiable current scientific 'fact'. And I for one would find such a graph immensely interesting and probably very useful.

 

[SteamKing]

Thanks, Harry - yes it helps. But I suppose I am looking for more something more rigorous than links between concepts. I'm interested in verifiable hypothesis dependencies - so if someone publishes a paper that, say, derives from other's works, those other works also have to be proven, and so on. So, if a hypothesis subsequently proves incorrect (new evidence?) then all works that are dependent also become incorrect, or at least, subject to re-proof. An audit trail, if you will.

It sounds to me like the scientific process should require it. As our knowledge grows, which often time is built upon other knowledge, we need to map this growth to be sure that all new knowledge is always founded on verifiable current scientific 'fact'. And I for one would find such a graph immensely interesting and probably very useful.

And who would be in charge of all this checking? And what is a "verifiable current scientific 'fact'" anyway? Obviously, you did not pick up on the hint about the great revolution in physics which occurred around the beginning of the TwenCen. Under your system, Einstein would still be working at the Swiss Patent Office and Planck would have gone on lecturing in Berlin. This idea smacks of having every hypothesis or theory checked by some central authority, much like the Church with which Galileo wrestled. This is a B-A-D idea.

 

[quitequick]

And who would be in charge of all this checking? And what is a "verifiable current scientific 'fact'" anyway? Obviously, you did not pick up on the hint about the great revolution in physics which occurred around the beginning of the TwenCen. Under your system, Einstein would still be working at the Swiss Patent Office and Planck would have gone on lecturing in Berlin. This idea smacks of having every hypothesis or theory checked by some central authority, much like the Church with which Galileo wrestled. This is a B-A-D idea.

Calm down, Steam! It's just an idea - not an attack on you, the scientific community or anyone else.

Checking? The same scientific community who currently peer review scientific papers, naturally.

"verifiable current scientific 'fact'"? Really? Ok - Those hypotheses that have been shown to be experimentally proven by multiple independent sources to some agreed high level of certainty (I think 5 sigmas is the threshold these days?).

The great revolution? Yes, I picked up on it. But ignored it because it was not relevant to my question.

I'm not sure how my "system" impinges on Einstein's or Planck's job prospects, as you suggest.

> This idea smacks of having every hypothesis or theory checked by some central authority
No. Your interpretation of the idea does this. I'm not sure how you made this leap from what I wrote.

 

[Evo]

Thanks, Harry - yes it helps. But I suppose I am looking for more something more rigorous than links between concepts. I'm interested in verifiable hypothesis dependencies - so if someone publishes a paper that, say, derives from other's works, those other works also have to be proven, and so on. So, if a hypothesis subsequently proves incorrect (new evidence?) then all works that are dependent also become incorrect, or at least, subject to re-proof. An audit trail, if you will.

A problem with the latest person's hypothesis would not necessarily have anything to do with someone's long ago original work. The new person might just be wrong. You don't think that citing the work of the people whose work you are building on is sufficient enough of a trail? I am not aware of this being a large problem, but you seem to be aware of many issues, and as a result are concerned that something should be done. Would you please post a specific example of a real life scenario where this was a problem because there was no idea which prior hypotheses or theories were wrong?

 

[quitequick]

A problem with the latest person's hypothesis would not necessarily have anything to do with someone's long ago original work. The new person might just be wrong.

My explanation must be unclear. I agree with your statement. The issue occurs when the long ago original work subsequently becomes subject to question - and therefore potentially invalidates the new person's work.

You don't think that citing the work of the people whose work you are building on is sufficient enough of a trail?

In itself, yes it is sufficient. But as knowledge upon knowledge grows it will become increasingly difficult to track the ever increasing line of dependencies. But even ignoring that potential issue, I think it would further bolster the veracity of a hypothesis through a more easily accessible audit trail of dependent hypotheses.

I am not aware of this being a large problem, but you seem to be aware of many issues, and as a result are concerned that something should be done.

No, I'm not - and I'm not sure why you would think I am aware of many issues.

Would you please post a specific example of a real life scenario where this was a problem because there was no idea which prior hypotheses or theories were wrong?

My thoughts were not prompted by some real life example. They were prompted by the notion that such an open and transparent process, which is the scientific process, would welcome any initiative that would make the 'knowledge tree/graph/hierarchy' immediately available and therefore, more easily provable. But more to the point, these knowledge dependencies are themselves, knowledge - or meta-knowledge, to coin a phrase. And if a knowledge dependency is potentially crucial to a whole arm of science which is built upon it, then having that dependency easily traceable to all dependent hypotheses, sounds important.

But if you insist and as an example only (and it is only an example), what if it became provable that the gravitational constant (G) was in fact not constant? What would be the impact on the sum of human scientific knowledge? What derived assumptions would no longer be correct? What hypotheses would be invalidated? I am sure a collection of individuals could maybe try and cover the sum of human scientific knowledge, but maybe they forgot a few things. Wouldn't it be nice to have a database to easily find these?

And imagine a searchable database, or a graphical representation of the proven connections between scientific knowledge. Sounds good to me.

 

[Evo]

I meant an actual example of where this has happened, not hypothetical questions. I don't know of any examples, so I was curious where you had run across this, and as you clarified, you haven't. That was all I was wondering.

 

[russ_watters]

My explanation must be unclear. I agree with your statement. The issue occurs when the long ago original work subsequently becomes subject to question - and therefore potentially invalidates the new person's work.

So to say what Evo said in a different way: I doubt that such an issue is common. I suspect it is extremely rare, if it is even possible.

This is one centralised database which captures the specific and required dependencies between proven scientific hypotheses. For example, hypothesis Z is only proven if hypotheses W, X, and Y are proven...
[separate post]

It sounds to me like the scientific process should require it. As our knowledge grows, which often time is built upon other knowledge, we need to map this growth to be sure that all new knowledge is always founded on verifiable current scientific 'fact'.

I think the basis here may be a misunderstanding of the scientific method. Theories aren't proven or dis-proven on the backs of previous theories/hypothesis, they are dependent on matching experimental data.

 

[flatmaster]

I think you may have better luck looking at inventions rather than theories. Every invention has prerequisite inventions that are required. These prerequisite inventions are more tangible and would be easier to identify than the prerequisite ideas needed for a new theory.

Reminds me of the technology tree in some games.

https://en.wikipedia.org/wiki/Technology_tree

 

[JesseC]

Isn't this basically what a citation web is? I think some websites can generate them (Web of knowledge?) linking a paper to all those it cites and all that cited it. If you took the links far enough I can imagine that you'd eventually get Pythagorus or something.

As other posters have pointed out, scientific study is not simply a set of dectuctions from one set of axioms. It is also a mess of inferences and paradigm shifts, so a purely dectuctive 'knowledge dependency graph' would not exist.

Edit: 'Citation mapping' http://wokinfo.com/products_tools/multidisciplinary/webofscience/citmap/

 

[robphy]

I'm not sure if this is relevant to your question
http://en.wikipedia.org/wiki/Automated_theorem_proving
http://www.cs.miami.edu/~tptp/OverviewOfATP.html

 

[Pythagorean]

quitequick: all scientific models are not a large, self-consistent web-work of theorems. There is no panacea: no one model (or network of models) of reality perfectly describes it. We have to move between models, often sacrificing generality for specificity (or vice versa). There's a lot of patchwork involved.

Even upon discovering quantum mechanics, we don't erase classical mechanics. There is no way to model cannonball behavior or the motion of the planets with QM.

That being said, there are several common themes and observations that apply to all areas of science, such as conservation laws. Newton's laws apply to the vast majority of classical observable phenomena... but there's still much ambiguity in specific systems as to how you intend to apply Newton's laws. Even within the same subject, different questions will require different aspects and focuses of modelling.