Scientific Insight after AI: Questioning the epistemic consequences of AI
|
Abstract
Traditionally, insight has been described as a sudden realization or discovery of a solution path that enables problem-solving. An ampliative inferential mechanism with key psychological elements like surprise and strong conviction are characteristic of insight (Cf. Peirce 1992, Aliseda 2006). Notably, such strong conviction often arises from the intuitive nature of the solution. While insights are commonly experienced in everyday life, they have been extensively studied for their pivotal role in scientific discovery and scientific knowledge. However, with the advancement of science and the emergence of new technologies, questions arise regarding whether our traditional understanding of insight remains applicable. Here, I discuss the epistemic consequences of AI implementation through the lens of the humanities, particularly from the philosophy of science. I focus on AI implementation’s role in insight formation mechanisms. I argue that learning from AI exposes human agents to novel problem-solving strategies, and when doing so, it demands an updated understanding of insights and the corresponding formation mechanisms. To do so I proceed in four steps. First, I characterize insight and provide a traditional example of insight in science. Second, I address some of the major changes AI has caused in how scientific knowledge is pursued and achieved and argue that this has an important effect on how insight occurs in current scientific contexts (Cf. Martinez-Ordaz 2023). Third, I argue that by learning from AI, human agents have slowly adopted new types of insight-formation strategies that privilege novelty, computational efficiency, and solution adequacy –over the intuitiveness of the solution. Here I defend that this demands an updated understanding of the epistemology of insights in AI-aided and AI-driven scientific contexts. Finally, I draw some conclusions. |
| ||
A methodological shift in favor of some paraconsistency in science
07-April-2022. 7th UNILOG
abstract
Many have contended that non-classical logicians have failed at providing evidence of paraconsistent logics being applicable in cases of inconsistency toleration in the sciences.
With this in mind, my main concern here is methodological. I aim at addressing the question of how should we study and explain cases of inconsistent science, using paraconsistent tools, without ruining into the most common methodological mistakes. My response is divided into two main parts: first, I provide some methodological guidance on how to approach cases of inconsistent science; and second, I focus on a peculiar type of formal methodologies for the scrutiny of inconsistent reasoning, the Paraconsistent Alternative Approach (henceforth, PAA) and argue that PAA can enhance a more accurate understanding of sensible reasoning in inconsistent contexts.
With this in mind, my main concern here is methodological. I aim at addressing the question of how should we study and explain cases of inconsistent science, using paraconsistent tools, without ruining into the most common methodological mistakes. My response is divided into two main parts: first, I provide some methodological guidance on how to approach cases of inconsistent science; and second, I focus on a peculiar type of formal methodologies for the scrutiny of inconsistent reasoning, the Paraconsistent Alternative Approach (henceforth, PAA) and argue that PAA can enhance a more accurate understanding of sensible reasoning in inconsistent contexts.
| Slides |
Ignorance of theoretical structure, scientific understanding, and the solar neutrino problem
|
4 June 2022
24 February 2022 |
abstract
Here, I contend that the ignorance associated with heavy technological implementation in physics is ignorance of theoretical structure and I explain how this ignorance relates to the production of specific insights and later on, to further epistemic achievements such as knowledge and understanding.
In order to do so, I proceed in four steps.
In order to do so, I proceed in four steps.
- First, I discuss the challenges associated with overcoming high degrees of ignorance, for different types of theoretical ignorance, in order to gain a deeper understanding of specific phenomena. I focus on those cases in which ignorance causes problems of the kinds of inconsistency and incompleteness.
- Second, I introduce a case study from neutrino physics: the anomaly in the measuring of the solar neutrino flux (cf. Bahcall and Davis 1976, Pinch 1986, Franklin 2000, Bahcall 2003, Martínez-Ordaz 2020). And I argue that the ignorance that caused the anomaly was ignorance of theoretical structure, significantly linked to heavy technological implementation; I contend that this ignorance limited the grasping of the inference patterns that held within the Standard Solar Model's predictions and the experimental results. Furthermore, I briefly discuss the questions of when did neutrino physicists start overcoming such ignorance and whether they have ever been in a sufficiently safe epistemic position to understand neutrinos in some significant way.
- Third, I address these questions and provide the generalities of an inferentialist account of scientific understanding that could be explanatory of the partial overcoming of ignorance of theoretical structure and the further achievement of knowledge and understanding --even if relevant high degrees of ignorance are still present.
- I illustrate this with the later evolution of the solar neutrino flux anomaly. Finally, I draw some conclusions regarding the bridges that connect ignorance of theoretical structure and scientific understanding in physics.
From Ignorance to Knowledge to Understanding in Big Data contexts
23-27 August 2021
|
|
|
Here I submit that the ignorance that emerges when using big data in the empirical sciences is ignorance of theoretical structure with reliable consequences and I explain how this ignorance relates to the production of specific insights and later on, to further epistemic achievements such as knowledge and understanding.
| ||
Comments to Jc's The contradictory christ
06 May 2021
|
|
This was a Author meets Critics session regarding JC Beall's recent book The Contradictory Christ.
The critics were Axel Barcelo, Luis Estrada-Gonzalez and myself.
My comments were published here.
The critics were Axel Barcelo, Luis Estrada-Gonzalez and myself.
My comments were published here.
Methodologies for the achievement of understanding in quantum mechanics: The case of primitive ontology
16 April 2021
VII International Workshop on Quantum Mechanics and Quantum Information: Quantum Ontology and Metaphysics
|
|
|
I argue that we can legitimately understand the standard Quantum theory and some of its interpretations, even if defective, if we can recognize the theory’s underlying inference pattern(s) and if we can reconstruct and explain what is going on in specific cases of the theories in question. To do so, I assume a Primitive Ontology methodology and I contend that it (remarkably) enhances our modal understanding of both the theories and the world that they describe.
| ||
From serendipity and ignorance to knowledge and understanding in big data practices
February 2021 SYMPOSIUM "Serendipity and Big Data"
|
|
| MAORSerendipity |
Can we use a contradictions right from the start methodology? / with Moisés Macías-Bustos
28 November 2020 Workshop Philosophical issues of quantum mechanics
|
|
for our talk, go to 1:50:00
abstract
Here we tackle the question under which circumstances, if any, should physicists adopt a methodology that accepts a contradictions right from the start in order to achieve better understanding of the quantum phenomena?
In what follows, we describe in detail what a contradictions right from the start methodology should be and we provide a general guide for adopting such a methodological approach in quantum physics.
In order to do so, we proceed as follows: First we introduce the contradictions right from the start methodology as it was presented by da Costa and de Ronde (2013), then we extend it into a more cohesive proposal of what this methodology should be in order to be relevant for the physicists’ practice. Second, we challenge the scope of this methodology by using it to tackle one of the most important problems of the GRW dynamical-collapse theory, namely, the ‘problem of tails’ (Albert and Loewer 1990, Wallace 2014). We reconstruct in terms of contradictions the problem of tails (in the form of both the problem of bare tails and the problem of structured tails). Third, we evaluate this methodology’s usefulness for explanation of this problem and contend that while the approach promises to enhance our explanatory power it’s not straightforward how it applies to these cases and whether it succeeds. Finally, we draw some remarks on the contexts in which physicists could (and should) adopt this methodology in order to achieve better understanding of the quantum phenomena.
In what follows, we describe in detail what a contradictions right from the start methodology should be and we provide a general guide for adopting such a methodological approach in quantum physics.
In order to do so, we proceed as follows: First we introduce the contradictions right from the start methodology as it was presented by da Costa and de Ronde (2013), then we extend it into a more cohesive proposal of what this methodology should be in order to be relevant for the physicists’ practice. Second, we challenge the scope of this methodology by using it to tackle one of the most important problems of the GRW dynamical-collapse theory, namely, the ‘problem of tails’ (Albert and Loewer 1990, Wallace 2014). We reconstruct in terms of contradictions the problem of tails (in the form of both the problem of bare tails and the problem of structured tails). Third, we evaluate this methodology’s usefulness for explanation of this problem and contend that while the approach promises to enhance our explanatory power it’s not straightforward how it applies to these cases and whether it succeeds. Finally, we draw some remarks on the contexts in which physicists could (and should) adopt this methodology in order to achieve better understanding of the quantum phenomena.
Studying disagreement to explain contradictions in the sciences
22 September 2020 Workshop Disagreement and Epistemic Injustice
In Spanish!
|
|
abstract
Here I address the benefits of analyzing the different types of peer disagreement that take place when a contradiction is discovered in a well accepted scientific theory. In particular, I contend that the scrutiny of the way in which such a disagreement might take place can shed light on three main issues: (i) the legitimacy of the contradiction that is at stake, (ii) its kind as well as (iii) the general procedures that can aid scientists to control the damage caused by the contradiction in question.
In order to do so, I proceed in four steps:
In order to do so, I proceed in four steps:
- First, I address the problem of peer disagreement and, briefly, sketch the four main standpoints on this issue: the Equal Weight View, the Steadfast View, the Justificationist View, and the Total Evidence View; I also present a typology of peer disagreements.
- Second, I discuss two historical case studies that illustrate both the discovery of a contradiction in a theory as well as peer disagreement with regard to the status of such a contradiction.
- Third, I explain how the analysis of these disagreements can be informative about the status of the contradiction as well as of the strategies that might be useful when working with it.
- Finally, I draw some conclusions on the value of peer disagreements for the understanding of contradictions in the sciences.
A structuralist approach to scientific understanding
28 July 2020 |
@Zoom Workshop “Understanding, in Science and Beyond”
abstract
Pace factualist and quasi-factualist approaches to scientific understanding, here I contend that scientists can achieve legitimate understanding of either a defective (vague, partial, conflicting, inconsistent, false) theory or an empirical domain via the use of defective resources. More importantly, I explain under which circumstances can defective elements be part of the content of understanding.
In particular, I argue that scientists can achieve such an understanding if they can recognize the theory’s underlying inference pattern(s) and if they can reconstruct and explain what is going on in specific cases of defective theories as well as consider what the theory would do if not-defective –even before finding ways of fixing it. To do so, I focus on cases in which the defective character of either entities or theoretical chunks might be both useful and an essencial feature of what is being understood. Therefore, the need of certain defects to be part of the content of understanding.
In addition, I submit that, when falsehoods (or any other defective elements) are included in the content of understanding, they must be joined by the inference patterns that allow them to remain well behaved.
-----------------------------------
∗This work is part of a larger project with Moisés Macías-Bustos in which we explore the possibility of achieving understanding of defective theories in the empirical sciences. See here
In particular, I argue that scientists can achieve such an understanding if they can recognize the theory’s underlying inference pattern(s) and if they can reconstruct and explain what is going on in specific cases of defective theories as well as consider what the theory would do if not-defective –even before finding ways of fixing it. To do so, I focus on cases in which the defective character of either entities or theoretical chunks might be both useful and an essencial feature of what is being understood. Therefore, the need of certain defects to be part of the content of understanding.
In addition, I submit that, when falsehoods (or any other defective elements) are included in the content of understanding, they must be joined by the inference patterns that allow them to remain well behaved.
-----------------------------------
∗This work is part of a larger project with Moisés Macías-Bustos in which we explore the possibility of achieving understanding of defective theories in the empirical sciences. See here
| |||
Understanding defective theories: From logic to epistemology
28 July 2020 Seminario de Lógica Iberoamericana (SeLoI)
|
|
abstract
It is undeniable that scientific understanding is a fundamental component of any successful scientific enterprise; understanding a theory allows scientists to find new domains of application for it, and understanding an empirical domain makes it possible to build new theoretical approaches to that domain. In addition, from an anti-exceptionalist view about logic, "logic isn’t special. Its theories are continuous with science; its method continuous with scientific method" (Hjortlan 2017: 1). In accordance to this, the epistemic aims of science should be also pursued (at least, partially) in logic. Being understanding one of the most valuable assets of the scientific enterprise, the achievement of it should also be an epistemic goal for logicians.
In light of the above, here I address two questions from the epistemology of logic, namely: can logicians achieve legitimate understanding of defective theories? and if so, how is this possible? I argue that logicians understand a defective theory if they can recognize the theory’s underlying inference pattern(s) and if they can reconstruct and explain what is going on in specific cases of defective theories as well as consider what the theory would do if not-defective –even before finding ways of fixing it.
If this is the case, the same recipe (or at least a very similar one) would explain how scientists from epirical disciplines can achieve understanding of both defective theories and empirical domains (via the reliable use of defective theories).
-----------------------------------
∗This work is part of a larger project with Moisés Macías-Bustos in which we explore the possibility of achieving understanding of defective theories in the empirical sciences. See here.
In light of the above, here I address two questions from the epistemology of logic, namely: can logicians achieve legitimate understanding of defective theories? and if so, how is this possible? I argue that logicians understand a defective theory if they can recognize the theory’s underlying inference pattern(s) and if they can reconstruct and explain what is going on in specific cases of defective theories as well as consider what the theory would do if not-defective –even before finding ways of fixing it.
If this is the case, the same recipe (or at least a very similar one) would explain how scientists from epirical disciplines can achieve understanding of both defective theories and empirical domains (via the reliable use of defective theories).
-----------------------------------
∗This work is part of a larger project with Moisés Macías-Bustos in which we explore the possibility of achieving understanding of defective theories in the empirical sciences. See here.
enhanced slides:
| UfLtE.pdf |
Is there anything special about the ignorance involved in Big Data practices?
18 June 2020 Colóquio Virtual SELF - Problemas Filosóficos
abstract
It is a fact that the larger the amount of defective (vague, partial, conflicting, inconsistent) information the more challenges scientists face when working with it. Here, I address the question of whether such challenges are of the same kind when working with ordinary-sized datasets than when working with Big Data. In order to respond to such a question, I focus on one particular epistemic challenge that comes naturally when dealing with large amounts of information, namely, ignorance. I submit that the ignorance that emerges when using Big Data in the empirical sciences is quite unique: it is a limit case of ignorance of theoretical structure with reliable consequences.
| SLIDESignorance_self |