Alexandra Karakas

To transform haphazard trial and error into enduring learning, memory is essential. Yet negative results, dead ends and missteps often go undocumented and are rarely shared formally in the journal archive. Still, such “negative knowledge” is often preserved in local contexts and informal communication networks, where it guides subsequent research through ad hoc catalogs of domain-specific missteps, or error repertoires (Mayo, Error and the growth of experimental knowledge. University of Chicago Press, Chicago, 1996). Here, I explore several historical cases and discuss how they function epistemically. Various forms of incongruences—discordant results, theoretical anomalies, and interpretive disagreements—yield new uncertainty and provide researchers a focus for further fruitful research through “troubleshooting,” isolating errors, and resolving the apparent discrepancies. However, when investigators find that a former justification was faulty, knowledge grows: a definitive error has occurred. Such errors are not useless, residual byproducts. Rather, knowledge of past error can help in interpreting unexpected new experimental results. Identifying possible sources of error also tends to raise evidential standards for subsequent research (at both observational and conceptual levels). Sometimes, they spur new methodologies to forestall or counterbalance errors. Error—and memory of specific sources of error and of general error types—ironically contributes to progress in science.

The underdetermination or ‘gap’ argument (GA), and the inductive risk or ‘error’ argument (EA), are the two main arguments used to defend the influence of non-epistemic values on scientific reasoning. However, they are often presented in a superficial or imprecise way, and their mutual relationship is not clear. This article analyzes their respective structures in detail, as well as their relationship with each other. The GA considers the logical structure of non-deductive inference (and of relative observation), and claims that (value-laden) background assumptions are needed to constitute (relative) observations and for those observations to confirm a hypothesis. It does not explicitly consider the consequences of these (value-laden) choices: rather, values are considered given and preexisting, so to speak. It is not normative: it claims that non-epistemic values are necessary to determine background assumptions in the sense that they are inevitable. By contrast, the EA considers the decision-theoretical problem of accepting or not a hypothesis, according to a (value-laden) required degree of confirmation. It is explicitly concerned with the consequences of these choices. It is normative: it claims that scientists should (in the sense of a moral obligation) take into account the non-epistemic consequences of their choices. By substituting the condition for hypothesis acceptance of the EA in the Bayesian account of the GA, one can provide a determinate limit when confirmation can be seen as sufficient to justify acceptance, thereby making the EA appear as a special case of the GA.

In this paper I point to different features and cases of scientific inquiry that suggest the relevance of a broad pattern of negativism. I have noted the role in categories of description and more normative epistemic and methodological standards across different sciences. One may call them shadow concepts and standards; they prove effective and puzzling. Finally, I have focused on causal description and reasoning and identified and examined important implications. The relational character of this dialectical pattern at work, synchronic and diachronic, and the cognitive, methodological and practical values it provides renders the picture not just plural, in a descriptive sense, but also pluralistic. This is contextual and integrated; in the case of scientific change, it is not the simple global dialectics of Popperian conjectures and refutations or Kuhnian paradigms and revolutions. The picture I provide for appreciating much scientific practice is one of contextual, synthetic pluralism that has been missed. Additional empirical examination can shed some light on the details of specific cases and their various forms of contributing to scientific change without substituting for a contextual account a monolithic Hegelian caricature.

This article provides a systematic and conceptual examination of failure in science. It first considers its marginalization in German academic discourse and its recent, albeit fragmented, resurgence. The author distinguishes failure from related concepts such as error and mistake, categorizes it in different knowledge contexts, and identifies its relevance for scientific practice. A model consisting of the three relata actor, goal and cause is developed as a philosophical tool and its analytical relevance is demonstrated using selected scientific examples. On a psychological level, failure is identified as an act of attribution (contrary to an objective fact) and, due to this interpretative character, can always be denied or even strategically avoided. Furthermore, failure is considered in science as a whole, in research and in teaching, whereby both its productive and its detrimental dimensions are identified. Contrary to its previous taboo, the article argues for a culture of failure in the sciences. In the quest for scientific knowledge, both the epistemic value and the potential damage of failure should be recognized.

My paper will explore the function of ignorance and error in the scientific research process. Scientific knowledge, as a particular form among many other forms of knowledge, can even be characterized by its attitude toward ignorance and error. The ground-laying epistemic feature of research consists in gaining knowledge about aspects of the world of which the actual knowledge is either scarce or absent altogether. “Specified ignorance,” as discussed by Robert Merton, is a state of affairs where the black spot of ignorance can be pinpointed. But research is also characterized by a kind of ignorance that we can address as “unspecified,” or “non-anticipated,” that is, by ignoring what one does not know. This often implies that what counts as knowledge today will reveal itself tomorrow as an error of the past. I will especially focus on the latter form of ignorance and present a few historical examples from the life sciences of the twentieth century.

When detecting scientific phenomena, researchers often aim to separate datasets into signal and noise. This signal-noise distinction captures that not all data one collects are relevant to this phenomenon. However, what happens when scientists “mistakenly” think that they have found a phenomenon of interest? In this chapter, I review the case of perytons, where terrestrial waveforms, initially of interest to radio telescope researchers, were unmasked as emanating from a break room microwave oven. This case exemplifies that the signal-noise distinction can be read in two different ways: that of data relevant to a phenomenon of interest and data irrelevant to it, and that of data with structure and data lacking it. I articulate these readings, discuss how they differently matter to detecting phenomena in datasets, and address how they complicate philosophical appeals to the signal-noise distinction. Further, I discuss whether this case counts as a mistake.

This paper examines the notion of malfunction in scientific instruments, focusing on the epistemic challenges posed by faulty performance. To frame these difficulties, I address the concepts of precision and accuracy as criteria for evaluating scientific instruments, showing how these notions intersect with, and sometimes obscure, the detection of malfunction. The discussion begins with a brief overview of relevant work in the history and philosophy of science on the philosophical problems of measurement. It then turns to the connections between malfunction, precision, and accuracy, highlighting the epistemic complications in their relationship. A case study of eudiometers, an eighteenth-century instrument developed to measure air quality, illustrates how theoretical assumptions, experimental practices, and shifting scientific frameworks further complicate judgments about precision, accuracy, and proper function. By examining how these devices could be mechanically precise yet epistemically misleading, the paper offers a richer understanding of the complexity of malfunction and the broader philosophical issues it raises for evaluating scientific instruments.

In this paper I propose that artworks may fail in various ways just like any other human endeavor—scientific, technological or humanistic. However, unlike other human projects that are defined by their a priori goals which provide criteria for determining their success or failure, art has no such external goals beside being a good artwork. Artistic goal is inherent to the work, but one cannot formulate any rule or principle by which art can be evaluated. Several types of artistic failure are presented: ugliness, meaninglessness, predictability, boredom, lack of originality, irrelevance or insignificance as well as kitsch. No truth theory can adequately account for such failures, yet it is undeniable that these types of failures partake in the artistic realm and express forms of artistic disappointment. I argue that we can understand the divide between artistic evaluation or, more generally, aesthetic evaluation, and any other type of evaluation, as a divide between two forms of order: aesthetic order that is indirectly measured without having any rules or principles to support such an assessment and discursive order that consists of external, a priori principles.

This chapter outlines three examples of scientific mistakes, two of which led to good outcomes and one of which, in my opinion, to a less than optimum outcome. I refer to the first as a case of ‘precocious pioneering’, the second (my own) as a case of ‘fortuitous stupidity’ and the third as a case of ‘mismanaged revelation’. The natural science in each case is gravitational wave detection, but the conclusions bear on science in general. The activity in which the mistake was embedded in the second two cases is not the natural science itself but social science in my case, and the interaction of science and society in the final case.

Whereas historical accounts typically depict the development of the Geiger Müller counter as culminating in the discovery by its inventors of ways to resolve problems of interference in 1928, the story of the continued development of the device in early cosmic ray physics has yet to receive sufficient attention. Far from representing a solution to the challenge of measuring cosmic ray radiation, the GM tube instead provided early researchers with a host of problems that demanded further innovations throughout the following decades. This chapter examines how the GM counter was transformed from an unstable and notoriously unreliable device to an indispensable element of cosmic ray physics. As we’ll see, achieving the levels of stability, durability and time resolution demanded by cosmic ray research required not only modifications to the device but also innovations beyond the walls of the counters themselves, which succeeded in drastically expanding the kinds of things GM counters could be used to achieve. As I’ll argue, it is only by considering the device within these wider contexts of equipment that we can understand the emergence of the GM counter as the central tool within the early cosmic ray physicists’ toolkit.

The common disease, common variant hypothesis (CDCV), still a fundamental tenant of genetic epidemiology more than two decades after it was first developed, may be defined as a fundamental set of assumptions about genetics and its importance to common disease (such as hypertension). The CDCV assumes that common ailments are caused by genetics that many human beings share, due to our shared evolutionary history. The CDCV, by focusing on common variation and its connection to health and disease, perturbed a century-long assumption that a significant number of conditions of medical importance were caused by single genes, and were relatively rare. This Mendelian emphasis in medicine has also translated to a Mendelian abundance in the historiography of genetics. This is the first detailed history of several topics around the history of common disease genetics. As importantly, this chapter uses the lens of the historiography and philosophy of error to assess and evaluate the CDCV. I find that the CDCV challenges notions of error and non-error in science. The CDCV, like the history of science and genetics generally, is not the story of truth overcoming errors, nor is it the tale of errors weighing down science and scientists. We can acknowledge the importance of the CDCV, while underscoring that rare variation and Mendelian conditions have their role to play. As importantly, it is unclear what could replace the CDCV or pose a better explanation. The opposing hypothesis around rare variation runs into related issues. In this way too, the CDCV poses considerable difficulties to established philosophies of science. The work of Karl Popper and his school, especially that of Paul Feyerabend, Joseph AgassiAgassi, Joseph and Imre Lakatos is discussed at length to provide a new perspective on recent genetics history, the importance of acknowledging errors in the history of science, and to aid in the assessment of the CDCV, which occupies a liminal space between truth and falsity.

The history of science is replete with debates surrounding the role of material culture in knowledge production, and philosophical discussions on epistemic and non-epistemic values in the sciences continue to flourish. Nevertheless, there are still underexplored issues at the intersection between the two fields, particularly regarding the value pluralism embedded in scientific instruments. This paper addresses key points from the gray area between these discourses, focusing on how epistemic and non-epistemic values influence the development and use of scientific instruments. Through a case study of the Venus Anatomica created by the sculptor Clemente Susini and anatomist Felice Fontana, this research explores the gendered representation of womanhood in eighteenth-century science. Using the values discourse framework, I explore the epistemic significance of material models while highlighting their value-laden nature through social, economic, and cultural factors. Although these models appear accurate at first sight, a value-based analysis combined with a close reading of the model’s lymphatic representation exposes their abstract and idealized portrayal of the female body. The paper also situates the Venus Anatomica in the historical context of anatomical teaching in Florence, Vienna, and Pest, and relates its visual logic to Paolo Mascagni’s representation of the lymphatic system. By combining philosophical analysis of epistemic and non-epistemic values with newly examined historical evidence about the circulation and educational use of these models, the paper shows how representational ideals of femininity shaped the epistemic use of anatomical wax models. It also argues that the omission of certain features, most notably the lymphatics of the breast, reflects both the historical limits of anatomical knowledge and the persistence of gendered representational conventions. This illustrates how non-epistemic values shaped Fontana and Susini’s representational choices, ultimately reflecting broader cultural and gender biases in the scientific practices of the period.

The Hubble Space Telescope (HST) stands as an iconic symbol of scientific exploration, its launch in April 1990 sparking great expectations. However, the initial images were disappointing, revealing blurred starlight and a failure to capture faint sources of light. NASA and optical technicians embarked on immediate troubleshooting, resulting in the discovery of various issues and disparate opinions on their root causes. This shows that defining the boundaries of an instrument’s intended function is complex, raising certain philosophical and practical questions. This chapter applies the philosophy of maintenance, in particular the notion of function and malfunction, to reveal epistemic issues surrounding faulty scientific instruments. The case of the HST highlights some of the epistemic problems concerning the maintenance of scientific instruments, and it also suggests that the relation between function and malfunction should not be seen as binary. Maintenance strategies hinge on implicit or explicit function definitions that are essential for understanding how artifacts operate. However, to delineate an artifact’s intended function, it is also necessary to understand when it is malfunctioning. The conventional view that an artifact is malfunctioning if it fails to fulfill its intended function is therefore insufficient. The case of the HST represents a partial malfunction somewhere between function and malfunction, demonstrating the intricate relation between science, technology, and maintenance. This approach sheds light on epistemic problems arising from flawed scientific instruments.

In contrast to the history of science and to science and technology studies, the value discourse in the philosophy of science has not provided a thorough analysis of the material culture of science. Instruments in science have a special characteristic, namely that they explicitly and clearly emerge from and remain embedded in social contexts, and are thus imbued with values. We argue that the materials (in most cases they are artifacts) used in science are necessarily influenced by both epistemic and non-epistemic considerations. A consequence of this is that a descriptive term cannot give sufficient information whether an artifact is performing in an acceptable way. Instead of the prevailing descriptive approach, we therefore advocate a normative notion of values in the material culture of science. To this end, we connect the material culture of science to the so-called “new demarcation problem”, in order to lay the foundations for a value-sensitive approach to the analysis of instruments. By assessing the five approaches of demarcation concerning value-influences, it will be shown that they break down at various points if the material aspects of science are taken seriously.