Open HyunkuKwon opened 3 years ago
I love Timmermans and Tavory's piece. Highlighting the importance of surprise in research and theory building seems like such a cool idea. I was left wondering how to apply some of these ideas in computational research. For instance, the idea of positionality, I've never seen it applied in computational or quantitative research; do you guys know of research doing it? Also, I believe Prof. Evans has transitioned over his career from more qualitative to computational work. I've always wondered how this trajectory evolved and how the qualitative origins of his work influenced how he thinks about computational research nowadays.
I am new to logic and I really like the three examples of white beans in this week's lecture which explain the similarity and difference of induction, deduction, and abduction. Timmermans, Stefan, and Iddo Tavory's paper introduces these three reasoning methods' role in data analysis to build a social theory. However, if I understand this paper right, the abductive research is creating new theory based on existing theory. So does this mean doing abductive research requires a wider knowledge of this area compared to inductive research?
Also, I am curious about the misuses of grounded theory--which in this paper are described as "the lackadaisical, incomplete, or inaccurate application of grounded theory principles", can you give us some examples? I think we should definitely avoid making these mistakes in our future research (or at least try our best to avoid it) .
I think this paper perfectly sums up where data and a potentially new theory could complement each other through close reflexive processes and being constantly aware of a researcher's role. Through the steps mentioned, the paper argues that ' the surprise, puzzle, or anomaly that may trigger a novel theory then emerges methodologically through careful data analysis against a background of cultivated theoretical expertise.' I am pretty much convinced, but at the same time I can't help but feel that this could be somewhat a lofty ideal, in that, if we are supposed to base our research on strong theories that come before, how does innovation really happen? at the same time, we are supposed to allow surprises and anomalies to reshape our prior understanding. There's is a reason for something to be a surprise to an academic field; what if we don't find anything to account for the surprises? when could we stop and say that we have a solid theory? In the end, I am just very curious about how well-received the abductive analysis is among social scientists.
The principles of qualitative research are deduction and induction. This paper introduces a new principle, namely abduction.
But what I wonder is how we should apply this idea. Unsupervised machine learning can help us humans to do induction reasoning. However, as the model becomes more flexible, its predictive accuracy becomes satisfactory, but at the same time its explanatory power decreases. I think we still tend to have better post hoc explanations of the problems we describe by simplifying the models. So, I would like to ask how we should strike a balance between the two?
The distinctions of different ways to construct theories provided a cool introductory reading in sociology for an outsider like me. My question is more about the direction of theory construction and is probably not totally related to this paper.
Interesting read! I found the examples comparing induction, deduction, and abduction through the relationships between entities A, B, and C especially useful. This reading led me to reflect on the methodological paradigms that CSS predominantly employ at this moment; in my experience so far, I felt that CSS can tend to lean towards 'blind' empiricism and feel disconnected from long-standing theory.
What examples of CSS research/paradigms do you feel are most representative of the abductive reasoning paradigm described in the paper?
I was wondering what challenges might abductive analysis face when, for example, we are proposing a new theory in art history with unsupervised methods? Would the "computationalization" pervasive in the social sciences persists in some "stubborn" area humanities?
This is an interesting epistemological paper discussing grounded theory and abduction and the differences between it and induction and deduction. My question is the following: Are practicing social scientists conscious about which method they follow when developing a social sciences theory and do they try to follow a specific epistemological pattern?
Similar discussions exist in natural sciences, but practicing scientists rarely care about such epistemological nuances (and most of them have never read Popper or Kunh or any philosopher of science); so I was wandering to what extend social scientists do take them into account when forming their theories.
Is there any risk that abduction, in privileging the element of "surprise" in an observation which makes us question underlying theory, risks de-emphasizing/ignoring/overlooking desirable lack of surprise? In other words, could abduction lead us to too readily challenge existing hypotheses or ways of explaining, which actually might fit the majority of the data available quite well?
What are some tips for getting a paper published with computational theory building? e.g. it seems hard to please both the typically qualitative theory-building peer reviewers as well as the quantitative theory-testing peer reviewers, both of whom may be assigned to your paper.
This is a really interesting paper talking about the essence of the theory construction. It argues that abduction but not induction should be the principle of empirically-based theory construction. Abduction reasoning, in other words "inference to the best explanation", seems to explain why those modern regression analysis methods(of course not limited only to regression) are popular and viewed as valid. I'm just curious about how it would link to the so-called over-fitting issues? Because abductive reasoning seems to prevail and one may always find data most suitable to his own "Theory". In this case, would he also modify a bit of his theory to fit the data and shall we blame him for that?
This article is quite informative! I have read some papers using computational analysis in a induction-deduction paradigm. For example, researchers may put a bunch of scraped data from social media into an algorithm and find that some conversational topics are the major ones that people try to avoid talking about. Then, they use survey data to validate their machine-generated hypotheses (e.g., people try to avoid specific topics in conversations). This is quite straightforward. So I am wondering what computational methods can do when abduction comes in. Thank you.
Abduction seems to be a less rigorous route than deduction and induction. As such, are there partial remedies for this potential problem?
Fascinating stuff. Suppose that our research result contradicts with our prior hypothesis, how do we know if the contradiction is because (1) our hypothesis is wrong, or (2) we operationalize our concepts in the wrong way?
I think the abductive analysis covered in this paper is very enlightening in creating innovative ideas in research especially when we are buried in the sea of data.
This seems like a sociology rendition of the "frequentist v.s. Bayesian", "inference v.s. prediction", "statistics v.s. machine learning." But I wonder what good does it make to overload those discussions with obscure theoretical terms, as the authors did in this paper. Maybe it aims at creating the next buzzword in sociological theory, but what novelty does it bring into the more general, cross-domain discussion?
Interesting paper! I really enjoyed reading about the logic of abduction and found the A,B,Cs useful. This paper was published in 2012 - it seems like a clean way to explain some relationships in data... how influential/used has this theory been?
This article touches on abduction as a form of positional knowledge. I’m curious about how positionality can be approached in a computational context, as I feel there’s a tendency for computational methods to obscure the socially situated nature of knowledge production. Are there any considerations you think are important in thinking through these ideas, or ways that you’ve seen positionality approached in computational research?
The idea of abduction requires the researcher to recognize surprising phenomena in light of existing theories. How does this idea interplay with Professor Evans' findings that newcomers to a scientific subfield have a higher likelihood of publishing novel and interesting research? Intuitively it feels like the existing theories can prime scholars' way of thinking about a field and an outsider with a fresh perspective and less familiarity with existing theory can think outside the box. Does this in some way clash with the notion of abduction?
What do you think about data-driven and theory-driven research?
What role does domain-knowledge play in Timmermans and Tavory's theory of abductive analysis? That is, how should a researcher balance the doctrinaire, dogmatic practices in their fields to draw novel insights about complex research questions?
The idea of reaching a conclusion after you've given data and existing theories time to marinate and synthesize through abductive analysis is rather creative. It does seem antithetical to computational methodologies, for which you must have a set procedure of steps into which you can plug in your data and output a conclusion. How might the idea of looking first at the state of the world and then coming up with explanations for it be implemented in computational models? Is there a way to infuse a less rigid, more creative essence into machine learning techniques?
It's an interesting paper on the abductive and inductive theories. I'm wondering how we can apply abductive theories in the more qualitative fields, like psychology. Personally I found it hard to distinguish these research processes in practice.
Post questions about the following orienting reading:
Timmermans, Stefan and Iddo Tavory. 2012. “Theory Construction in Qualitative Research: From Grounded Theory to Abductive Analysis.” Sociological Theory 30(3) 167–186.