Post APtgOaWKH6LIpgEX2m by jacomyma@mas.to
(DIR) More posts by jacomyma@mas.to
(DIR) Post #APtVR3knSRPma7ygqm by tiago@social.skewed.de
2022-11-23T11:46:38Z
0 likes, 0 repeats
Periodic reminder that you should not use modularity maximization to find community structure in networks.https://skewed.de/tiago/blog/modularity-harmful#networkscience #datascience #modularity @networkscience @complexsystems
(DIR) Post #APtVboFfQzplgjXkQq by jboy@post.lurk.org
2022-11-23T11:48:32Z
0 likes, 0 repeats
@tiago hey @jacomyma, time to bring the controversy mapping to the fedi
(DIR) Post #APtVyJFCsxLmBH3ZtA by tiago@social.skewed.de
2022-11-23T11:52:40Z
0 likes, 0 repeats
@jboy @jacomyma is it really a controversy when arguments to in only one direction? 😉
(DIR) Post #APtWDBNwO6Am9MEZVY by jboy@post.lurk.org
2022-11-23T11:55:19Z
0 likes, 0 repeats
@tiago @jacomyma ok, let me start then! :eyeless_triumph: I would hold that "purely descriptive" is only a problem when seeking answers to some kinds of research questions, not others. RQs in the humanities and social sciences frequently do not need inferential methods.
(DIR) Post #APtWjkzfX10Rr2Pi7c by tiago@social.skewed.de
2022-11-23T12:01:13Z
0 likes, 0 repeats
@jboy @jacomyma The post was about the fact that modularity maximization should not be used. To argue in the opposite direction you should say why it should be used. I'm not aware of such an argument.As I wrote below, indeed there are situations where descriptive approaches are desired, not inferential ones. I even list some of them. So this cannot possibly be a reason for controversy.https://skewed.de/tiago/blog/descriptive-inferential
(DIR) Post #APtXXzRK5zPZ8ZIlYe by tiago@social.skewed.de
2022-11-23T12:10:17Z
0 likes, 0 repeats
@jboy @jacomyma You say many research questions in the humanities and social sciences require descriptive, not inferential methods. I'm curious. Can you give an example?
(DIR) Post #APtZwcnxaXuW5HlShk by jboy@post.lurk.org
2022-11-23T12:37:08Z
0 likes, 0 repeats
@tiago @jacomyma Fair enough. I don't think I can make a strong case FOR modularity maximization. I suppose it has the strength of being easily interpretable. Not sure to what extent that is true of SBM.Making a stronger case would require me to have a grasp on its competitors. As somebody who *applies* network techniques, my grasp doesn't reach far enough.So I do what is normal under the scholarly division of labor and go with what is established in my field and practiced by people whose work I trust to be good.When I see you arguing against that, I have to assess whether the arguments you are making against established practice should disqualify it for my uses, or whether you are coming from someplace very different in making this judgement.(Sorry, not good controversy fodder I fear.)
(DIR) Post #APtaZ7Ufw5nYGi6GXo by jboy@post.lurk.org
2022-11-23T12:44:05Z
0 likes, 0 repeats
@tiago @jacomyma well for instance a starting question may often be purely descriptive (what is going on?), and a follow-up question may be interpretative (what does it mean?).For instance, one may first ask what kinds of subgroups appear to exist, and then ask how differently positioned people perceive these social divisions. Neither of those questions needs to get at the underlying process that created the apparent subgroups.
(DIR) Post #APtbECLsmJ9DD0Zhx2 by tiago@social.skewed.de
2022-11-23T12:51:29Z
0 likes, 0 repeats
@jboy @jacomyma when you say that modularity is more interpretable than the SBM, my head spins. I really wonder what you mean by this.Regarding the “popularity” advice: when I was very small, my parents taught me that if everyone else jumps off a cliff, I should not do the same. I think it's good advice.Applied scientists should have their own methods to evaluate their tools. Otherwise, how do you know that it works?
(DIR) Post #APtbXGZDKa8Rmdu2pk by tiago@social.skewed.de
2022-11-23T12:55:00Z
0 likes, 0 repeats
@jboy @jacomyma so if the nodes are connecting to each other entirely at random, and your method finds a bunch of interesting looking groups, you're OK with that?
(DIR) Post #APtbj23v5Kr8GysMPA by jacomyma@mas.to
2022-11-23T12:48:39Z
0 likes, 0 repeats
@jboy @tiago It's a controversy if the disagreement persists. It's the case here. Other clue: the participants instrumentalize the controversy. For instance by saying that it goes only one way. :-)But damn guys, you're doing the controversy without me! I'm in a conference in Budapest, I was literally in the audience answering you when one of the speakers asked a question to me... Now I need time to catch up!
(DIR) Post #APtbj2tfyuFwrV5h4q by tiago@social.skewed.de
2022-11-23T12:57:06Z
0 likes, 0 repeats
@jacomyma @jboy my point is that there is no disagreement, hence there cannot be a controversy.
(DIR) Post #APtc1alcn97r7YAVBg by jboy@post.lurk.org
2022-11-23T13:00:26Z
0 likes, 0 repeats
@tiago the "nodes" connected by human action, which is never random nor entirely accounted for by a single mechanism. @jacomyma
(DIR) Post #APtcBcFY8N3AcBwfY0 by tiago@social.skewed.de
2022-11-23T13:02:16Z
0 likes, 0 repeats
@jacomyma @jboy Also, I'm not “instrumentalizing”, it's a statement of fact: no one proposes arguments *in favor* of using modularity.I only hear conservative, reactionary positions based on inertia, “everyone else does it”, etc.
(DIR) Post #APtckaP4zSrR1Z4tOa by tiago@social.skewed.de
2022-11-23T13:08:36Z
0 likes, 0 repeats
@jboy @jacomyma You're evading my question. I didn't make an assumption about the real world, I postulated a hypothetical scenario, a counterfactual. *IF* your nodes connect completely at random, and the method gives you rich community structure, would this outcome be a problem for you?
(DIR) Post #APtd6JLn4dasVQtmmO by pintoch@mamot.fr
2022-11-23T13:12:29Z
0 likes, 0 repeats
@tiago @jacomyma @jboy oh I am so glad you are bringing this up again, it reminds me of the good times in Copenhagen! I am grabbing some popcorn and send very dense cliques of hugs to you all!
(DIR) Post #APtdrD2ANNmMWg5AvI by jboy@post.lurk.org
2022-11-23T13:20:58Z
0 likes, 0 repeats
@tiago @jacomyma Not evading at all. As requested, I was trying to give an example of how RQs in the social sciences may differ. Another difference is that your proposed counterfactual doesn't make sense because humans (we!) don't work that way.I understand the problems you are talking about regarding modularity. I am just trying to understand whether you are being hyperbolic.
(DIR) Post #APtenRlhkfCj1J9W64 by tiago@social.skewed.de
2022-11-23T13:31:30Z
0 likes, 0 repeats
@jboy @jacomyma Of course you are evading, you refused to answer my question twice already!The point of the counterfactual is not to make a statement about reality, but to expose what are the underlying criteria that are being used.
(DIR) Post #APtfbadj2h8R0sKXnU by jacomyma@mas.to
2022-11-23T13:34:29Z
0 likes, 0 repeats
@jboy @tiago I would remove any if(false) statement from my code. Why should I care about what's after the "if" when I know it's false?
(DIR) Post #APtfbbCowDCwlikZV2 by tiago@social.skewed.de
2022-11-23T13:40:35Z
0 likes, 0 repeats
@jacomyma @jboy Because you want your theories to be falsifiable.And also because you may want to illustrate a point, or expose how a logic is used.“If a dinosaur comes running at you would you be afraid?”“I refuse to answer the question because this situation is impossible.”
(DIR) Post #APtg1DFYcyxW04FBEf by tiago@social.skewed.de
2022-11-23T13:45:13Z
0 likes, 0 repeats
@jacomyma @jboy My examole about random graphs is important because it illustrates that people do interpret community detection in an inferential way, despite what they claim.The fact that you refuse to answer the question does little to convince me otherwise.
(DIR) Post #APtgNPSZuayAEuv0ZE by tiago@social.skewed.de
2022-11-23T13:49:13Z
0 likes, 0 repeats
@jacomyma @jboy But let me indulge you a little bit. Suppose we do an experiment. We arrange two sets of people. I set A people behave as they wish. In set B, the control group, we ask the participants to use a random number generator to choose friends. We want to compare A and B, but our algorithm finds communities in both of them. Is this a problem?(And please don't refuse to answer and say that you would never do that this experiment or whatever.)
(DIR) Post #APtgOaWKH6LIpgEX2m by jacomyma@mas.to
2022-11-23T13:49:27Z
0 likes, 0 repeats
@tiago @jboy I'm not in the business of convincing you; that's a mean not a goal. I just want to explore the discussion.The argument in favor of modularity is: it fits people's needs better. I may not last forever. It does not mean that inferential methods could not be better, or that they are not better in some situations. They actually are. And still, for many people modularity is strictly better. That's the argument I think.
(DIR) Post #APtgu9Hn4EWxdJU3aS by jacomyma@mas.to
2022-11-23T13:55:08Z
0 likes, 0 repeats
@tiago @jboy I am precisely refusing to answer because yada yada. Seriously.I think we've agreed already that in this situation inferential methods make more sense. The disagreement branches out upstream.
(DIR) Post #APtgvwGYRiUppcl8ZU by tiago@social.skewed.de
2022-11-23T13:55:29Z
0 likes, 0 repeats
@jacomyma @jboy how can you make this claim without saying what people's needs are?If their need is just to draw a pretty picture, that's one thing. If it's to draw inferential conclusions, that's another.We need context to ground this discussion.
(DIR) Post #APthIMrqDuEghmtXfc by tiago@social.skewed.de
2022-11-23T13:59:30Z
0 likes, 0 repeats
@jacomyma @jboy Then why stall for so long?Anyway, I don't think @jboy agrees with you... And we haven't gotten to the bottom of the question if there are indeed many questions in the social sciences that are descriptive, not inferential.
(DIR) Post #APthRX1joreEoxtSxE by jacomyma@mas.to
2022-11-23T14:01:11Z
0 likes, 0 repeats
@tiago @jboy I want to catch you writing black on white that you cannot accept that people may have other goals than those you see, and that non-inferential methods are invalid whatever they are.Or, equally interesting, that non-inferential methods may be better for some goals. And then we ask if they exist and what they might be.
(DIR) Post #APtjhzLKVbSdBQ24Su by jboy@post.lurk.org
2022-11-23T14:26:32Z
0 likes, 0 repeats
@tiago @jacomyma sure, happy to say that would be a problem. It also sounds like a very uninteresting study :)
(DIR) Post #APtk23RixljrhUak8u by tiago@social.skewed.de
2022-11-23T14:30:11Z
0 likes, 0 repeats
@jacomyma @jboy I have no idea why you want to corner me in this straw-man, since I have already explicitly refuted it.Why not argue with what I say, instead of what you think I believe in secret?Not only have I recognized that it all depends on the question asked, but I have made two categories for them.In the descriptive category, I have identified two *legitimate* clustering goals: graph-partitioning for optimization (VLSI, task scheduling), and characterization of random walks on networks.I didn't put a third example because I could not find one. That's why I always ask people who claim that not all interesting questions are inferential what exactly do they mean — I'm seeking that elusive third example. I'm sure there must be something out there!I also believe, as I state explicitly, that *most* interesting questions related to community detection are inferential, despite what many people claim. This is why I made the random graph litmus test, which I think helps revealing this bias, since tasks often fail it.Note that the latter is only a belief, not a dogma, and I am always interested in updating that belief based on new empirical information.
(DIR) Post #APtkVwLPvoiKMWtN4a by vtraag@mastodon.online
2022-11-23T13:43:05Z
0 likes, 0 repeats
@jacomyma @jboy @tiago In general, I would tend to agree with @tiago that we could be more careful about inference. Having said that, the purported mechanism in an SBM *only* concerns graph generation, not any process that takes place on top of it. In the prototypical example of @tiago, in the random graph (bottom-right), there might well be an opinion formation process, leading to an outcome where node's opinions would cluster, as can be seen in the "community" network (bottom-left).
(DIR) Post #APtkVx4REQj6c9xKFM by vtraag@mastodon.online
2022-11-23T13:45:32Z
0 likes, 0 repeats
@jacomyma @jboy @tiago to make matters more complicated, typically, in most social systems, processes take place simultaneously (i.e. processes *on* graphs and graph generation processes). This is how I would interpret the comment by @jboy that it is almost never accounted for by a single mechanism. In such cases, the results from a particular clustering algorithm might be informative, even if there is no clear inference about the graph generation process.
(DIR) Post #APtkVxgiw5LqWtrtvE by vtraag@mastodon.online
2022-11-23T13:47:46Z
0 likes, 0 repeats
@jacomyma @jboy @tiago at the same time, as said, we could be more careful about our inference. This would mean trying to be more specific about graph generation processes versus graph dynamical processes. In many cases, such inferences are doomed to fail however (e.g. general confounding between homophily and social influence), but in some cases we could perhaps infer some aspects of certain processes.
(DIR) Post #APtkVyCd1Ss87qnNeS by vtraag@mastodon.online
2022-11-23T13:58:44Z
0 likes, 0 repeats
@jacomyma @jboy @tiago Finally (or probably not, expecting some comments ;)), I think in many cases, people are not necessarily interested in inferring the exact process of network generation and network dynamics. For instance, in this picture by @manlius, he observes that in the bottom-right (green) group there are many complexity scientists, which seems to be interesting. https://mathstodon.xyz/@manlius/109376457129435179 I'm not sure which method he used, but that's the point, it remains interesting.
(DIR) Post #APtkVyg3G4PLb6YsVs by manlius@mathstodon.xyz
2022-11-23T14:10:25Z
0 likes, 0 repeats
@vtraag @jacomyma @jboy @tiago I admit it: for the viz I have tried the Leiden algorithm, the SBM and the Louvain. Eventually I opted for the latter because it found less clusters, but my intention was just to convey that the complex systems group is there (and the result was the same across 3 methods).
(DIR) Post #APtkVzFr6x31O9JTJw by tiago@social.skewed.de
2022-11-23T14:35:32Z
0 likes, 0 repeats
@manlius @vtraag @jacomyma @jboy FFS Manlio. Et tu brute.
(DIR) Post #APtkxUxlKSFZY3y6yG by jacomyma@mas.to
2022-11-23T14:40:34Z
0 likes, 0 repeats
@tiago @jboy I do that because my opinion is not formed. I am just listening. But although I have stated multiple times the third example, you never acknowledge it. Let me try again.People need colors that are sufficiently aligned with the layout so that they can talk about it. They do not use it to explain, nor they pretend to. Yet it's part of their scientific process. A totally legit process (for ex. leading to rigorous, empirically grounded, falsifiable claims).Or am I missing something?
(DIR) Post #APtlecU7ErNbKSfC0O by tiago@social.skewed.de
2022-11-23T14:48:17Z
0 likes, 0 repeats
@vtraag @jacomyma @jboy @manlius What you highlight is precisely the distinction I make between inferential and descriptive methods.A method that describes random walks on a network is not inferential about its structure.This is well captured by the litmus test I proposed: if Infomap finds clusters in fully random graphs, it is not actually overfitting, since it's not *fitting* in the first place. The random walk really gets trapped in those groups.Now, very often people are perturbed by this litmus test, even when they use infomap, etc. This often indicates that what they are really after is an inferential statement. I think this more often the case than not.Regarding your comment on @manlius's figure: how do you know the community found is interesting, knowing that the method used can find those communities in random graphs?My guess is that you are using the network layout as additional information. But then in this case, why do we need the damn community detection method in the first place? And what happens when the network is too big to draw? And what if the drawing gives you fake structure too (as it most definitely does)?By all means use whatever you want to do quick and dirty stuff that's between you and your rabbi — but don't sell it for more than it is.
(DIR) Post #APtm5v6a7HVOtjfns0 by vtraag@mastodon.online
2022-11-23T14:53:16Z
0 likes, 0 repeats
@tiago @jacomyma @jboy @manlius the limitation, I believe, is that you interpret "inferential" only with respect to the graph generation process. However, it might be equally valid to try to make inferences about dynamical processes on graphs that result in what we would consider clusters. This is different from description, and is about inference.
(DIR) Post #APtm8hwzTvcwy5W5RI by tiago@social.skewed.de
2022-11-23T14:53:49Z
0 likes, 0 repeats
@jacomyma @jboy It does not seem to me like a well defined task, but maybe I just need to understand it more.In order for it to qualify in my framework, it needs to pass the litmus test: If you see clusters in a random network, that align with the 2D layout, is this a problem for the analysis or not?You say that they need the clusters to “talk” about them? What do they talk about? Isn't this crucial?
(DIR) Post #APtmV3dPXbVCaj11LU by tiago@social.skewed.de
2022-11-23T14:57:51Z
0 likes, 0 repeats
@vtraag @jacomyma @jboy @manlius I don't interpret it like this, I *define* it like this, because this difference is crucial.In my text I state explicitly that infomap is making an inference about a hypothetical random walk given a network — a parameter of this process. It's just not making an inference about the network structure (the actual data)It *is* legitimate to make an inference about a hypothetical dynamics on a network.What is not legitimate is confuse the two.
(DIR) Post #APtmaa36S7GQZRfMsC by jacomyma@mas.to
2022-11-23T14:58:49Z
0 likes, 0 repeats
@tiago @jboy My understanding of what you're saying is that if this need does not fit your networks, it is not legit? Or it does not exist?Hypothesis: the need I talk about is so mundane that it does not fit your world view.That need exists for people who do not try to model. They just need a temporary way to write about the network map, to build statements such as "look at the green cluster in the bottom-right". They need to make intersubjective statements.
(DIR) Post #APtmqaovLuwfb1fZ8S by boogheta@piaille.fr
2022-11-23T15:01:42Z
0 likes, 0 repeats
@tiago From a personal empirical view, all I can say is that the very unstable results produced by the SBM inferential approach (for instance 15 different clustering hierarchies over 100 runs) in comparison with modularity classes only jittering a few nodes from one community to the other makes me often way more confident in the latter. A fortiori when some ground knowledge of the studied field confirms obvious gross descriptive communities.That's not mathematical, just purely empirical.
(DIR) Post #APtmvHVUaRJD91oekS by manlius@mathstodon.xyz
2022-11-23T15:02:22Z
0 likes, 0 repeats
@tiago @vtraag @jacomyma @jboy I agree. The point of that figure was just to have a representation, without any inferential request. I would not put that figure in a research paper: I would use the degree-corrected SBM for the analysis or, at best, explicitly state that the colors code some group structure without pretending this to be correct.
(DIR) Post #APtn7rDjd8ZDNpVUMS by vtraag@mastodon.online
2022-11-23T15:04:50Z
0 likes, 0 repeats
@tiago @jacomyma @jboy @manlius Yes, this is how you define it indeed. But that is exactly the problem. You limit the concept to fit (no pun intended ;) exactly the SBM model, namely that it is about the graph generative process. However, "inference" as a term understood by most people will not be limited to the graph generation process.
(DIR) Post #APtnHtpUmiLyN3xQZc by tiago@social.skewed.de
2022-11-23T15:06:41Z
0 likes, 0 repeats
@jacomyma @jboy No, something is only illegitimate if it's used in a manner that is not consistent with it's goals. So I want to understand the goal, hence the two categories, litmus test, etc.I completely understand that there are people who do not want to do anything sophisticated. They want to draw a network, look at some clusters, and say things like: look, this big cluster only talks to this other big cluster, etc.However, I think that if you would take these same people aside and say: look, actually, these groups are all the outcome of random fluctuations, like monkeys typing on a keyboard. I do think that many people would understand that that it was a statistical illusion, and be somewhat disappointed in their visualization. Don't you think so?
(DIR) Post #APtnyCiQxOFH6jpKEK by tiago@social.skewed.de
2022-11-23T15:14:18Z
0 likes, 0 repeats
@vtraag @jacomyma @jboy @manlius First of all, I don't limit anything to the SBM, this is silly, You can clearly fit a latent space model, a growth model, a model with triadic closure, etc. This is not about the SBM.I also don't "limit" anything to the graph generative process, I just make a distinction when this is done and when it's not done — which frankly, is totally confused in the literature.And finally, traditionally the idea of fitting/inference is something you do to data. This is not my idiosyncratic view, it's overwhelmingly standard. And the data is the network, not the random walk that you imagine. This idea of fitting a process that you imagine is pretty non-standard, and I haven't seen anywhere else, actually.
(DIR) Post #APtoP4JYIiFas0zykK by jacomyma@mas.to
2022-11-23T15:19:08Z
0 likes, 0 repeats
@tiago @jboy It's not about being unsophisticated, it's about having different goals.Making intersubjective statements is not an unsophisticated way of doing modelling. It's its own thing.Not everything relates to randomly typing monkeys.
(DIR) Post #APtoo6I85THeF2Gl04 by tiago@social.skewed.de
2022-11-23T15:23:41Z
0 likes, 0 repeats
@booghetaI see, so you want a method that is *confident* in its results. For you, confidence is a proxy for correctness?
(DIR) Post #APtotnuHM0XuxZphyK by boogheta@piaille.fr
2022-11-23T15:24:42Z
0 likes, 0 repeats
@tiago No my main point is I want a method that provides consistent results across multiple runs.
(DIR) Post #APtp0ftdXD9c8woD7A by tiago@social.skewed.de
2022-11-23T15:25:53Z
0 likes, 0 repeats
@jacomyma @jboy I used “unsophisticated“ not in a demeaning way, only in a manner that does not involve going deeply into technical details and theory, which is reasonable for an applied scientist.
(DIR) Post #APtpLlCH1ResfEa37Y by boogheta@piaille.fr
2022-11-23T15:26:21Z
0 likes, 0 repeats
@tiago So far all my experiments with Graphtool's implementation of the SBM failed to do that, which is a little confusing to me and the researchers I'm working with I must admit.
(DIR) Post #APtpLllMuxjOQ504p6 by tiago@social.skewed.de
2022-11-23T15:29:44Z
0 likes, 0 repeats
@boogheta Giving the same answer every time is what being confident means.Think about this: what if these multiple results tell you something different but equally valid about the network, because it can be explained in different ways?And what if the single way that modularity clusters is actually an overfit, and reproduces mostly noise?I wrote a paper about consensus and dissensus in network clustering that you might find interesting:https://journals.aps.org/prx/abstract/10.1103/PhysRevX.11.021003
(DIR) Post #APtpSRwL1AUsxzJgOm by jacomyma@mas.to
2022-11-23T15:07:20Z
0 likes, 0 repeats
@manlius @tiago @vtraag @jboy My attempt at rephrasing what you say: you need to build an intersubjective statement where what you write relates to the picture in a way that the reader can go and look for themselves. And the details do not matter much, for ex. the exact boundaries of the green cluster, precisely because you expect the reader to go look at it. At no point is there a notion of modelling involved.Do you agree with this description?
(DIR) Post #APtpSSXYmmGspQjPPs by manlius@mathstodon.xyz
2022-11-23T15:11:09Z
0 likes, 0 repeats
@jacomyma @tiago @vtraag @jboy that figure is art. Shape and colors are there to convey the message that there is a structure.No modeling or inference involved, in that figure. For inference etc, what I think it's clearly stated in our latest paper:https://www.nature.com/articles/s41467-022-34267-9
(DIR) Post #APtpSSy9BvXS9tAdrE by jacomyma@mas.to
2022-11-23T15:23:59Z
0 likes, 0 repeats
@manlius @tiago @vtraag @jboy Ah, I liked my description better but eh, too bad for me.It looks like you say "art" to emphasize that the statement carries no explanatory powers. Which is true, but has nothing to do with art.Making intersubjective statements is a very real need in science, if mundane. Not all the science process is about claims with explanatory powers. It's still science. Science is sometimes mundane.
(DIR) Post #APtpSTMbiz6XNkcAz2 by tiago@social.skewed.de
2022-11-23T15:30:56Z
0 likes, 0 repeats
@jacomyma @manlius @vtraag @jboy Can you clarify what you mean by “intersubjective”?
(DIR) Post #APtpx1N7qYttXpQKkC by tiago@social.skewed.de
2022-11-23T15:36:30Z
0 likes, 0 repeats
@jacomyma @jboy “Not everything relates to randomly typing monkeys.”I'm sorry (or happy) to say, information theory underlies more than what you may think.
(DIR) Post #APtq6fhhGOg1EneCpc by jacomyma@mas.to
2022-11-23T14:14:43Z
0 likes, 0 repeats
@manlius @vtraag @jboy @tiago Excellent example, I confirm. I do observe that people care *a lot* about the number of clusters, because too many of them may defeat the point of having them in the first place.
(DIR) Post #APtq6gDbLmCIpkZgYq by tiago@social.skewed.de
2022-11-23T15:38:14Z
0 likes, 0 repeats
@jacomyma @manlius @vtraag @jboy Some people care *a lot* about crack, but this does not mean you should give it to them.
(DIR) Post #APtqF8cXdRUzEqH0BU by jacomyma@mas.to
2022-11-23T15:39:46Z
0 likes, 0 repeats
@tiago @manlius @vtraag @jboy The term loaded in history of science, but in short: a statement appealing on the writer and the reader having the same subjective experience of something.That thing is therefore simple enough to guarantee it in practice.One could say that all scientific facts ultimately rely on intersubjective statements, because we have to agree to read the same measurements on the apparatus.
(DIR) Post #APtqJIAo3oc21XHUm0 by pintoch@mamot.fr
2022-11-23T15:40:27Z
0 likes, 0 repeats
@tiago @jacomyma @manlius @vtraag @jboy I think I should be getting more than popcorn at this stage 😄
(DIR) Post #APtqre8etsVFzger8y by jacomyma@mas.to
2022-11-23T15:46:42Z
0 likes, 0 repeats
@tiago @jboy Ah, at last we are on the verge of the kind of methodologically imperialist argument I think your approach can lead to.Not all gatekeeping is bad, but I will not follow you on the ground where you allow yourself to think you know what other people are doing better than themselves.Let's call it a personal preference.
(DIR) Post #APtrVR54rsUF4yxKZE by tiago@social.skewed.de
2022-11-23T15:53:56Z
0 likes, 0 repeats
@jacomyma @jboy So, if I say that you cannot understand networks without mathematics, or literature without words, I am an "imperialist" now?Do I really need to point out that flat rejection of absolutism is in itself an absolutist view, etc?Anyway, what I dislike the most about this discussion is how easily it floats to the rarefied air of general platitudes, and away from concrete examples, use cases, formal frameworks, etc.
(DIR) Post #APtroFDDDhoJlrjI80 by jboy@post.lurk.org
2022-11-23T15:57:18Z
0 likes, 0 repeats
@tiago @jacomyma I agree that if people jumped to the conclusion "look, this green cluster is there because they are caught in a filter bubble" or whatever, that would be a problem.But if the clusters are a starting point for further inquiry drawing on additional methods and data, which they very often are, then their imperfections do not doom the whole enterprise.
(DIR) Post #APtsBMw6fUBYO4ksYS by tiago@social.skewed.de
2022-11-23T16:01:30Z
0 likes, 0 repeats
@jboy @jacomyma Sure, in end it's about what standards you want to set.I think a scientific field is only as good as its standards.
(DIR) Post #APtsOyjXRxcdNVMl8q by jacomyma@mas.to
2022-11-23T16:03:58Z
0 likes, 0 repeats
@tiago @jboy Well, at least give me that I did not essentialize you. I did not say you were an imperialist, nor do I think you are one.I am merely disclosing what my problem is with that line of reasoning. It's not the journey, it's where it leads to. As I said, my opinion is not formed. I genuinely want to see if this line has another destination. It may. And if not, then maybe that's just how it is.Thank you for the discussion.