<img attachment="metrology.jpg" align="left">The article <a href="https://ieeexplore.ieee.org/abstract/document/9908261" author="Luca Mari; Dario Petri" source="IEEE">Measurement, Dataism and Post-Truth Ideology: The Good, The Bad and The Ugly</a> is quite an interesting description of the state of philosophy vis à vis data science.<fn> You can find the original, in Italian, at <a href="https://issuu.com/tutto_misure/docs/tm.4-2021" author="" source="Issuu">Tutto_Misure n.4 - 2021</a> on pages 103 and 104. Or you can just <a href="{att_link}tm.4-2021_mari_it.pdf">download a PDF of that article here</a>. <bq>Dataism, at least in its most radical position, sees the universe as a gigantic computation system, whose state transitions are in fact "embedded" computations, so that empirical phenomen are actually streams of data and living organisms are biochemical data processing systems [6]. Hence data are the objective representation of reality, or even the reality as such, and therefore all that is required to understand the empirical world and to make appropriate decisions on it.</bq> <bq>A positive and constructive understanding of this position, that is not the trivially self-contradictory "everything is relative," might be devised from the consideration that no available evidence, however specific it is, provides a certain support to a unique theory. Rather, different and incompatible theories may account for the available data.</bq> <bq>Starting from the acknowledgment that absolute truth is not accessible to us, post-truth advocates post-truth make the radical move to claim that there is no truth at all. According to them, opinions dominate on facts and emotions prevail on rationality. Or perhaps, even more radically, post-truth ideology assumes that there is nothing like facts free of interpretation, and that rationality is only a superstructure of a particular culture developed in a particular historical period: scientific knowledge would only be a contingent manifestation of a societal organization that has promoted competition through research.</bq> So deconstructionism is dispatched as an adjunct of fake news, and all of Quine's research and philosophy on <i>qualia</i> is swept away as relevant only for generating more conspiracy theories. The conclusion of to which considering cultural context inevitably leads---according to this paper---is that <iq>scientific evidence is denied as such</iq>, which is making the strawman do a tremendous amount of work. <bq>reliability is only a matter of goodness of data fitting. But for dataists this is a further reason of advancement: they would claim that causality is an unclear, unscientific concept [91, and removing it from technical endeavors should be considered a significant step forward.</bq> The dataist way lies madness, wherein we simply trust whatever output we get because the process couldn't be biased in any way that we can see. And, thus, we end up with candidate pools curiously bereft of women and POC, but we dare not wonder whether context or culture could have had anything to do with that, lest we be branded as heretical post-truthers, a mode of thought that used to be associated with those concerned about how we think we know what we know, like Quine (and his qualia), Derrida, or Foucault. Instead, they are are a lumped in with Trump and his coterie of self-obsessed hangers-on, who are more interested in attention and power than in getting at what is truth. Dataists glibly assume that there is a single truth that can be proven, irrespective of context, despite all evidence to the contrary. We just have to try harder, I suppose. Either try harder to shore up our modelsget more data (dataists no longer believe models are necessary...this would be admitting that their data expresses anything less than the absolute truth)...or try harder at ignoring the gaping holes in them. And definitely don't consider it unfair to elect Trump or any other information-poor ideologue---of any political bent; they are not constrained only to the right wing---as the leader of the post-truthers instead of dealing with their concerns honestly. <bq>This is the role traditionally played by causal explanations, that are not only highly efficient data compression devices (compare the number of bits required to store a formula representing a physical law with the size of a dataset from which the same information provided by the formula can be inferred by regression), but also tools able to provide socially understandable and criticizable justifications for the decisions that are made.</bq> I find it interesting that this paper leads with such heavy-handed and extreme definitions of both dataism and so-called post-truth-ism. It's hard to believe that anyone seriously interested in improving our use of information holds these radical positions. I would not be surprised in the least to find that the most powerful people in the world, in charge of pulling the levers that govern billions of lives, are not all interested in what is actually true, but what is personally beneficial to them. Should their personal interest somehow overlap with what is societally beneficial, well, then, win-win. But it's not necessary. So, while yes, there are enough people in these extreme camps, and, yes, they are powerful, they are also not going to be reading this paper, nor are they likely to change their behavior when scientists point out to them that they're not behaving rationally or in the interests of either their constituents or <i>the truth</i>. <bq>The distinction between data, information, and knowledge is crucial here: the "deluge of data" is now a fact, but it does not imply that there is also a deluge of information, and even less of knowledge.</bq> <bq>[...] the Weaver's interpretation of Shannon's information theory, in which communication problems are understood as being constituted of three "levels" of subproblems:<ul> "Level A (the <i>technical</i> problem): how accurately can the symbols of communication be transmitted? Level B (the <i>semantic</i> problem): how precisely do the transmitted symbols convey the desired meaning? Level C (the <i>effectiveness</i> problem): how effectively does the received meaning affect conduct in the desired way?" </ul>This interpretation may be applied and generalized to problems of data (and information, and knowledge) treatment, in terms of:<ul> a <i>syntactic</i> (i.e. "level A") problem: are data <i>formally correct</i> (and therefore, in particular, clean, not missing, ...)? This question focuses on data as syntactic entities, only able to be identified as different from each other (zeroes need only to remain different from ones); a <i>semantic</i> (i.e., "level B") problem: is the information obtained from data <i>representative</i>? This question assumes that information is data that refer to something and therefore to which a meaning has been assigned; a <i>pragmatic</i> (i.e., "level C") problem: is the knowledge obtained from information <i>useful</i>? This question assumes that knowledge is useful information in a context, provided that knowledge is justified true belief.</ul></bq> <bq>Indeed the position of the upper surface of the alcohol in the capillary has a value in metres (or millimeters), not in kelvins (or degrees Celsius). This position value (i.e., the data) is mapped to a temperature value (i.e., the information) by means of the instrument calibration: without the knowledge provided by the calibration (e.g., in the form of a calibration curve), indication values cannot be related to measurand values.</bq> <bq>[...] metrologists are well aware that it is critical to decide when it is not valid anymore (and therefore, when the instrument needs to be recalibrated), a situation that in Data Science corresponds to assessing whether the data-generating process is not stationary anymore.</bq> <bq>the metrological culture can foster the transition from an information society, characterized by pervasiveness of new technologies, to a knowledge society, in which information is a crucial factor of social development. While an information society only generates and disseminates raw data and information, a knowledge society transforms the widely available information in knowledge useful to support effective decision making and the improvement of human conditions.</bq> <hr> <ft>I have the original PDF, but I'll be damned if I upload a 9MB file for four pages of text. IEEE have outdone themselves in producing a document that defies nearly all compression.</ft>