3. Discovering Higher-Level Patterns - oritenting

[JunsolKim]

Post questions here for this week's oritenting readings: Timmermans, Stefan and Iddo Tavory. 2012. “Theory Construction in Qualitative Research: From Grounded Theory to Abductive Analysis.” Sociological Theory 30(3) 167–186.

[isaduan]

I really enjoy the idea posed by the authors. Abduction depends crucially on the researcher's theoretical sensitivity - I suppose 'theoretical sensitivity' requires years or decades of training; how can we, as junior researchers, grow our theoretical sensitivity? Does this suggest that we should ground ourselves deeply into a subfield/ a narrow theoretical literature, as early as possible, and extend ourselves elsewhere only after we are grounded somewhere? I am personally still jumping around different theoretical fields and methods and having the trouble of picking one ground!

[Hongkai040]

I really enjoy the idea posed by the authors. Abduction depends crucially on the researcher's theoretical sensitivity - I suppose 'theoretical sensitivity' requires years or decades of training; how can we, as junior researchers, grow our theoretical sensitivity? Does this suggest that we should ground ourselves deeply into a subfield/ a narrow theoretical literature, as early as possible, and extend ourselves elsewhere only after we are grounded somewhere? I am personally still jumping around different theoretical fields and methods and having the trouble of picking one ground!

I love the idea of abductive analysis too! My concern of abduction requires a broad familiarity with existing theories is different from Isabella's. I'm not sure if I know the clear definition of 'existing theories', especially at our time. I hold the belief of science as a whole. Many problems in one field may be addressed by borrowing ideas and theories from other disciplines. It's not always possible to get ourselves familiar with extensive existing theories from candidate disciplines. If not feasible, does that mean other means like interdisciplinary collaborations are more important?

[Sirius2713]

I really enjoy the idea posed by the authors. Abduction depends crucially on the researcher's theoretical sensitivity - I suppose 'theoretical sensitivity' requires years or decades of training; how can we, as junior researchers, grow our theoretical sensitivity? Does this suggest that we should ground ourselves deeply into a subfield/ a narrow theoretical literature, as early as possible, and extend ourselves elsewhere only after we are grounded somewhere? I am personally still jumping around different theoretical fields and methods and having the trouble of picking one ground!

I love the idea of abductive analysis too! My concern of abduction requires a broad familiarity with existing theories is different from Isabella's. I'm not sure if I know the clear definition of 'existing theories', especially at our time. I hold the belief of science as a whole. Many problems in one field may be addressed by borrowing ideas and theories from other disciplines. It's not always possible to get ourselves familiar with extensive existing theories from candidate disciplines. If not feasible, does that mean other means like interdisciplinary collaborations are more important?

Adding on my two classmates' comments, I also like the idea of abductive analysis. I think that's what many empirical researches are doing by building on literature review and extracting new ideas from new data. And my question is how can we make sure the novelty of our new theory rather than just some slight changes to an existing theory? Because we now have massive amount of data and theories, it's easy to get something new with slight changes instead of genuinely constructing a new theory.

[pranathiiyer]

Like everyone, I understand and appreciate the idea of abductive analysis, and I agree with what my classmates have to say! Moreover, I think a lot of times research can also be "unsurprising" in the sense that it might be using actual empirical evidence to validate something that intuitively makes sense or an existing theory that has its roots in more theoretical contexts. This could also extend to studies which are similar to already conducted studies, but just look at a different population under consideration. Would such studies be "less abductive" then? Perhaps there is no answer that would be agnostic of context, but would like to know what others think.

[GabeNicholson]

In the article the authors mention, "Peirce saw an innate tendency toward a positive truth." And then later call this assumption "High problematic". But evolutionary psychology has shown overwhelming evidence that humans are evolved problem solvers with specific evolved mechanisms designed to locate the truth (Argumentation, gossip, causal inference, etc.) Similarly, the authors then mention that "Abductive analysis, consequently, rests for a large part on the scope and sophistication of the theoretical background a researcher brings to research." But this also isn't completely true, and it's most obvious when looking at prolific researchers. Take John Von Nuemman for instance, who posted seminal papers in his field before the age of 25—far from having a large theoretical background, he was simply a gifted intellect that could reason far faster and better than everyone else. Clearly, innate talent isn't everything, but it is far more important than the authors give it credit.

[Jiayu-Kang]

Like everyone, I understand and appreciate the idea of abductive analysis, and I agree with what my classmates have to say! Moreover, I think a lot of times research can also be "unsurprising" in the sense that it might be using actual empirical evidence to validate something that intuitively makes sense or an existing theory that has its roots in more theoretical contexts. This could also extend to studies which are similar to already conducted studies, but just look at a different population under consideration. Would such studies be "less abductive" then? Perhaps there is no answer that would be agnostic of context, but would like to know what others think.

I also had the same concerns! But when I'm trying to think further on this, I realize it's difficult to define based on whether what the study tries to validate "intuitively makes sense." When I see a social theory I often have a feeling of "isn't it obvious?" and I think that's probably because a theory became a theory since it at least makes some intuitive sense (especially in the field of sociology). Then, can we say some studies are "less abductive" because they only validate something we all think should be true? Or it this just our subjective sense of "obviousness"?

[ValAlvernUChic]

I didn't expect - but love immensely - to see the influences of phenomenological writings (Heidegger, Husserl, etc.) in this paper! The authors mention the concept "ready-to-hand", in Heideggerean speak literally to perceive something as a tool, but there was no mention of the "present-at-hand", to see the world as a scientist/observer, which I think would've quite nicely fit with the detaching/disassociative principles of defamiliarization. Importantly, the paper itself elaborates on and showcases abduction performed through very largely ethnographic methods, where the researchers lived in the context and the process of data collection (e.g. being in the hospital to listen to the nurses). For us, this process of collection and analysis seems a lot more distanced from the actual contextual embeddedness from where the data was collected which, at least from my reading, might make our "rose-tinted lenses" through which we analyse our data colored with theories more so than an intimate understanding of the social world we are interested in. I was wondering whether there is an implication that by studying the theory, we can fill this gap? If so, what else does that imply? If not, does it matter?

[sizhenf]

I like the authors' argument on using empirical data and methods as part of the abduction analysis in social science research. In particular, the idea of trying to fit data with existing theory and what doesn't fit could inspire new theories. I believe a lot of research ideas actually do come from this process. It also helps avoid the issue of "cherry-picking", in some sense. If we only use data to serve the purpose of verifying a theory, then we could selectively choose the data points that fit better with our argument.

My concern is there seems to be some constraints regarding when the abduction analysis could be used. Essentially, we would need to have data of very good quality in order for it to play such an important role. However, in many fields of social sciences, this is not a satisfied condition. Either we have very small dataset, or there's missing data issue. I wonder what the authors would say about the use of empirical analysis in these kind of situations.

[MengChenC]

To me, the author argued in a way of abductive analysis being constructed and leveraged from the cultivated theoretical sensitivity and methodological heuristics. From this perspective, the power of abductive analysis is promising and convincing. However, this also reminds me to reconsider its limitations, and if these two approaches are mutually exclusive. That being said, we lack a unified and overall theory dominating various scenarios in many domains, such as in social sciences. So in what context we can unify these two mechanisms and further concrete the theory we are looking at?

[melody1126]

The authors point out an important factor to abductive analysis: "the ability to recognize a finding as surprising in light of existing theories." (page 169) The theoretical relevance of our research is just as important as finding a good data source. How would we train for this ability to recognize surprising qualities in our data and analysis?

[mikepackard415]

I'm curious how the concepts of induction vs. abduction for theory generation map onto the concept of the "burden of knowledge." I think this relates to Isabella's question up above. It is harder to know what is surprising when you have less experience in a field, but I don't think we would want to limit theoretical contributions to only the most established scientists. How do we strike a balance here?

[facundosuenzo]

I really enjoy the idea posed by the authors. Abduction depends crucially on the researcher's theoretical sensitivity - I suppose 'theoretical sensitivity' requires years or decades of training; how can we, as junior researchers, grow our theoretical sensitivity? Does this suggest that we should ground ourselves deeply into a subfield/ a narrow theoretical literature, as early as possible, and extend ourselves elsewhere only after we are grounded somewhere? I am personally still jumping around different theoretical fields and methods and having the trouble of picking one ground!

That's a good point. An opposite interpretation (and in tension) for this theoretical sensitivity could be grounded in the fact that young scholars may be prone to have "fresh eyes" towards data that may favor an abductive approach. That doesn't necessarily mean not knowing, but embracing creativity and innovation while keeping a "free-floating attention" as proposed Freud in psychoanalysis.

[Jasmine97Huang]

i particularly appreciate the importance of acknowledging one's positionality when conducting abductive analysis (or any type of scientific inquiry, in my opinion). However, it doesn't seem to be the case in real life where we don't see a lot of researchers elucidate their positionalities in their work. The way we conceptualize and identify cultural anomaly largely depends on our experiences as unique individuals with blindspots and prejudice. I think it is a challenge for everyone, especially social scientists, to acknowledge the perspectives we bring into our analysis of the complex social world.

[Qiuyu-Li]

How do abduction, induction, and deduction associate with our common practice of using data to conduct causal inference? I think I got lost in the story of logic, and I may be just requiring an example to understand the paper.

[hshi420]

The abductive analysis is not specific to text as data methods, but more like a fundamental methods for social science or science qualitative study theory construction. How can we incorporate this in our social science analysis using text as data methods? In other words, aside from theory construction, in which stages can we use abductive analysis?

[konratp]

On page 180, the authors, referring to other forms of qualitative analysis, contend that proponents of "such approaches [...] purposefully eschew methodological "cookbooks," worried that they would result in a fetishization of methods and crass empiricism. Instead, they direct researchers to privileged theoretical data points or to specific heuristics for specifying and amending theory."

To me, it sounds like neither approach is ideal. Of course, one could (like the authors) argue that finding a balance between the two is crucial and a "right" way of conducting research. But what if instead of taking the both of best worlds, utilizing an abductive approach instead worsens the negative aspects of each approach? Are abductive analysts at large not running a greater risk of being ignorant in some form given the greater number of opportunities for ignorance to creep into the process?

[NaiyuJ]

I find this article very very insightful for young social science researchers. The paper is more about theory construction, and our class is more about a range of advanced methodologies. Computational methods, like machine learning, do not only serve as new empirical strategies in social science research but more importantly, even change the way how we theorize our observations. I would like to hear more about how we can use machine learning to adopt this kind of abductive analysis and how we can use the text-as-data methods to construct better theories!

[YileC928]

I appreciate the idea of ‘abductive reasoning’, and as many of my classmates above, I also have this long-existing confusion (perhaps started from high school) that what is the contribution of a study if it just validates a priori ideas that an ordinary person could reason intuitively.

Besides, I am interested in the phenomenon – the division of labor between theorists and empiricists. It seems that not only in sociology, but in many other disciplines such as economics, we tend to label researchers as either theorists or empiricists, and few people focus on both. Could this be some sort of evolution that naturally happens to increase the aggregate efficiency of research?

[chuqingzhao]

I really enjoy reading this paper about abductive reasoning and it provides inspiring thoughts of creative production of hypothesis.

One concerns is that as the abductive analysis a basically qualitative data analysis and it requires revisiting and defamiliarizating the phenomenon, I wonder whether and how abductive analysis also can be applied to quantitative analysis? And if possible, what is the difference or similarity between abductive analysis in quantitative studies and qualitative studies? For example, developing measurement and conceptualizing theoretical notion can be critical in quantitative analysis. In some cases, different measurements could generate different results. How to justify or falsify different measurements in order to develop new theories?

Additionally, I love the point brought by @YileC928. She mentioned the division of labor between theorists and empiricists. To add on her points, I am personally curious the paradigm shift in social science studies-- why researchers have embraced quantitative analysis from social science research to policy making? How can we form effective communication between theorists and empiricists?

[LuZhang0128]

I've read a few papers on a similar topic but with an opposite perspective. I am a very theory person yet find myself loving this article as well. I wonder, however, if this abduction fits sociology research as well as it claims? We know that data from social science can be messy. Most of us may have spent quarters studying potential biases. I wonder if we will eventually reach a wrong conclusion through this approach.

[chentian418]

it's my first time reading paper about abductive reasoning, And I find the process of creatively inferencing and double-checking these inferences with more data really appealing. Abductive reasoning also utilize the data by moving back and forth between data and theory iteratively, which is more grounded in both sides than induction. While I find induction is used more often in the field of economics, I was wondering what are the major impediment for economists to seeks this kind of situational fit between observed facts and rules? I am also curious about, to what extend can we perceive a phenomenon as related to other observations in the sense that there is a hidden cause and effect relationship?

[zixu12]

It seems to me that we are doing abductive analysis when applying the tools for big data in this course. Or is it more like inductive anlysis? As this paper is based on sociological discipline, I am wondering whether this kind of analysis is generally workable in other disciplines as well.

[Emily-fyeh]

In the idea of abductive reasoning, I think researchers can move away from the typical inductive/deductive framework and look at the whole picture of tools and theories. In social science disciplines, each sub-fields has distinct practice on how to construct the grounded theories and how to subsequently explore the knowledge system. Therefore, the point of abductive reasoning might be providing a way to position the new piece of phenomenon-interpretation pairs.

[kelseywu99]

An interesting read on abduction and its logical ground. As the authors argue that abduction requires researchers' extensive familiarity with broad theoretical backgrounds who recognize the "potential relevance" of unanticipated observations, I was wondering what exactly makes this process different from the inductive analysis? I assume the process of induction requires one's knowledge in order to grasp surprising findings to generate new theory, too, as discussed in the first few chapters of our textbook?

[sudhamshow]

The concept of abductive learning seems a little 'abstract' to me. At its core, isn't it similar to grounded theory? What are the significant differences from grounded theory? How is 'surprising research evidence' even measured?

While adductive learning does seem to produce more novel and innovative theory, researchers have often found success with the most generic and simplest explanations sometimes. The theory of Occam's razor which is often cited alongside the application of decision trees or ensemble learning has seemed to generalise well without over-fitting. Is this a contradiction to what abductive learning proposes?

[ttsujikawa]

The idea of abduction is pretty interesting and makes me understand two sections of research: an inductive process that should be more creative and flexible, and a deductive process that should be more critical and evidence-based. Considering the core of the abductive research process, I was wondering that if intuitively some existing theory does not make sense, tackling the theory with rigorous deductive research could be considered as an abductive process?

[ZacharyHinds]

Abduction is pretty interesting, and I agree with the notion of its importance to social research. While there could be conscerns about "seek[ing] theory" instead of facts, I think it is a given that it social research is inherently complex given the immense complexity of the social world. What I wonder, and struggle with, is how to best apply this sort of reasoning to our own analysis and research? They cover it somewhat in the text, but more concrete and actionable ideas might help in grappling with it.