3. Clustering & Topic Modeling to Discover Higher-Order Patterns of Meaning -orienting

[lkcao]

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.

[XiaotongCui]

Very interesting research! While the paper emphasizes the need for researchers to have substantial theoretical and methodological training for Abductive Analysis, is there a risk of excessive reliance on such training, leading researchers to become too entrenched in existing theoretical frameworks during actual research, lacking flexibility and innovation? Could this result in overlooking or dismissing new perspectives or findings that deviate from established theories?

[sborislo]

I can certainly see the benefits of abductive analysis! However, it seems less efficient, due to the inherent openness to new theories based on anomalies in the data. Already, in psychological and behavioral research, people take shortcuts to provide a cohesive (even if unempirical) paper. How is abductive analysis feasible in a profession where most researchers are incentivized to argue for a singular theory (usually a "pet theory," as Timmermans and Tavory call it) in the shortest time possible? In other words, how can abductive analysis be worth the cost to—more than likely—having to investigate multiple chains of theoretical explanations, if people's incentives are not aligned with finding the best set of explanations for a phenomenon?

[chanteriam]

The authors stress the qualitative nature of abduction and its use in theory generation. How will the use of NLP and other computational tools in qualitative analysis influence these popular methods of research (induction, deduction, abduction) and, more specifically, popular theory construction practices as they rely less on the researcher needing to thoroughly familiarize themself with the data?

[cty20010831]

According to the authors, one important role of abductive analysis roles is pinpointing surprising parts of the findings that challenge or extend existing theories, which reminds me of the different ways to reconcile theoretical disputes in social science. One example that come right out of my mind is moderation, which is often used by social scientists to find the balancing points between conflicting theories. Hence, I am wondering does finding boundary conditions a form of abductive analysis?

[Dededon]

I'm interested about how could we broaden our range of computational RQs with the logic of adbuction analysis. The logic of abductive analysis, described in this paper, is to first revisit a common phenomenon, finding the alternative case through analysis, and reintegrate the new outlier case back into the original theory. Computational methods have their strengths in discovery of patterns in the step one of revisiting the phenomenon, and step two of detecting the outlier. Yet, many social phenomenons are bound by data constraint for the intervention of computational analysis. How are we going to break free of that?

[Twilight233333]

I'm actually curious if we can do factory production theory? In my opinion, a theory is an abstraction and generalization of a universal thing. If a theory is obtained by stimulating abductive reasoning through the process of re-examining, defamiliarization and substituting cases according to the theoretical knowledge suggested by the author, is what we actually get a theory that can summarize the law or a story that explains the data? Does factory production make the theory less explanatory and end up with a story?

[Caojie2001]

The abductive analysis proposed in the paper is definitely a novel concept about grounded theory that reveals a possibility of improvement based on the existing disadvantage of traditional grounded theory, which is based purely on induction. However, the conceptualization part still seems ambiguous to me. For example, revisiting of the collected data is proposed as an effective way to sensitize different theoretical approaches. However, I think that it's almost unavoidable that the data itself is collected under certain theoretical frameworks that tend to fit the data to some existing theories. Thus, there seems to be a tension between intentionally collected data and the generation of new theories to a certain extent. How do we deal with this tension in actual research work?

[joylin0209]

The article talks about the combination of qualitative analysis and computational tools. Does this lead to thinking about the interaction between technology and humanities? With the continuous development of data analysis tools and the impact of humanities research methods, how to find an appropriate balance between theory and data causality?

[Vindmn1234]

This paper advocates for a shift from purely inductive methods towards an abductive approach in social science research. My question is: How can NLP facilitate the analysis of large-scale qualitative data sets in a way that complements and enhances Grounded Theory and Abductive Analysis methodologies?

[chenyt16]

This article proposed abductive analysis as a qualitative data analysis to generate "creative and novel theoretical insights." I have two questions regarding implementing it practically. (1) The abductive method adds emphasis on generating a creative hypothesis based on "surprising evidence," but how do we define "surprising"? If an existing theory can explain half of the observed evidence, should we code it as "surprising," generate a creative new hypothesis, or should we seek adaptations to the existing theory? The author admitted that "abduction rests on the ability to recognize a finding as surprising in light of existing theories and presumes in-depth familiarity with a broad range of theories.". Since social science is a highly multidisciplinary field, this characteristic of the abductive method places a very high standard on researchers. A surprising piece of evidence in one field could be explained by theories in another field. For example, prospect theory uses psychological effects to illustrate economic phenomena. (2) The same piece of evidence may have different hypotheses for explanation, and sometimes, the same evidence can even be used to support opposing viewpoints. For example, the phenomenon of linguistic recursion can either be used to support or challenge universal grammar. Moreover, if one excessively focuses on the evidence at hand and derives new theories based on it, it may make it challenging for these theories to converge and may lead to fragmentation.

[YucanLei]

It is very interesting how this paper talks about the novel "abductive analysis approach" relative to the main stream "inductive approach". However, I still think abductive approach is rather far reached. This idea mainly comes from the fact that social science research is somewhat different from the natural science. In natural science, the facts exist and is waiting to be uncovered whereas for social science research, it is possible for us to actually create the outcome we want. We might be able to "lure" the results we would like to see. How can we prevent this from happening?

[bucketteOfIvy]

Timmermans and Tavory (2012) propose an abductive framework, and provide a few goals that methods should achieve. One of these goals, defamiliarization, is accomplished in their framework largely by the movement from atextual data to textual data, with the process of transcription and coding inherently defamiliarizing the data. However, when doing content analysis, our data is largely textual as is. What other nuts and bolts methods can we use to defamiliarize our datasets?

[ana-yurt]

In Timmermans and Tavory's article (2012), the method of abduction requires a trained eye to look for unanticipated findings and ways of modifying existing theories. I am curious what are examples of defamiliarization for computational textual data, which relies very much on existing on-the-shelf tools for analysis. How does topic modeling relate to the practice of abduction?

[Marugannwg]

From my first impression, this paper seems to sell the idea of applying abductive analysis in research cycles to encourage a kind of leap of faith in exploring new theories. Indeed, abduction is qualitative -- you are creating something new based on the sole/limited anomaly you find in data that could not be explained by previous grounding theories. This is putting your explanation at risk (which sounds intimidating), but also allows the community to compare, improve, and discard different theories and progress as a whole.

A small concern arises as the abductive process seems to be driven by anomalies or surprises... If social scientists are expected and encouraged to look for surprises, they might want to create and tinker towards them, making a lot of mess... Is such a "surprise" something to be foreseen in the research planning stage? What is a reasonable expectation to set?

[yueqil2]

Social Scientists always spout off causal influence and highlight the deductive reasoning, but this paper draws us back to alternative type of reasonings-inductive reasoning and adductive reasoning. Even though our perception of the phenomenon driven by theoretical formulations will easily fall within the realm of existing theories, the lighthouse for research is still needed. How can we code without theory? Field notes and coding are also based on theoretical mechanism. Otherwise, what's the difference between them and anthropological records? The point is how careful the coding should be for adductive analysis. Is there a standard?

[QIXIN-ACT]

As someone who has delved into management research during my undergraduate studies, I've noticed a strong focus on theoretical frameworks. This emphasis often leads to a reluctance in accepting new theories. How do we effectively evaluate and compare these new theories with existing ones? When should we adhere to established theories, and when should we pivot to innovative ones?

[Audacity88]

The authors note that their abductive framework requires an understanding of the theories which predominate in a field. They claim, however, that this understanding will not lead to mere deduction in the case of sociology because sociology is not a "paradigmatic science" in which a single theory dominates, but one in which multiple theories proliferate. I wonder then if the abductive approach can work for more paradigmatic sciences or if sociology is an exception?

[yuzhouw313]

Abductive analysis demonstrates a strong theoretical robustness through its dual engagement with existing theory and the methodological rigor applied in uncovering novel theoretical insights. As Timmermans and Tavory suggest, this approach enables the extension and revision of existing theories by intertwining data collection, analysis, and theory development in a recursive manner. However, this prompts a critical question: At what point do we determine that an existing theory is partially flawed and in need of revision, or when is it necessary to develop an entirely new theory? Moreover, the abductive process could potentially yield an overwhelming number of theories, potentially complicating the search and review of existing literature. How might we address this challenge in the practice of abductive research?

[erikaz1]

The concept and practice of abduction is exciting. Abduction, as the authors write, is not new (likely something everyone practices in every day life), though less often discussed compared to inductive and deductive reasoning in an academic context in my experience. What I was most drawn to was the positioning of the researcher in order for one to more intentionally practice adductive analysis. Timmermans and Tavory write that the researcher ideally has a sophisticated, deep, and broad theoretical background, to which they can compare evidence or surprising ideas.

1. What exactly is the nature of this theoretical background? In order to be surprised by novel observations related to those very fields or invite a multiplicity of interpretations from empirical evidence, wouldn't we need to simultaneously have theoretical familiarity and hold room to doubt that which we have become familiar with?

2. Also, In trying to understand abduction in relation to deductive and inductive reasoning, are we overemphasizing its ability to consider "surprises" and "outliers"? Maybe the crux of the ideology is its recursive "process of double fitting data and theories".

[anzhichen1999]

It is very exciting to see Iddo Tavory‘s aricles in our reading list this week. I have taken two of his classes back in the undergrad (Social Theory and Quantitative Methods in Social Science)! In the article, the authors discuss defamiliarization as a crucial step in abductive analysis. Could you provide specific examples or case studies where defamiliarization significantly altered the initial interpretation of data and led to groundbreaking theoretical insights? Additionally, how do you recommend researchers cultivate the skill of defamiliarization in their analytical processes?"

[ethanjkoz]

Coming from a qualitative background in sociology, I feel as though I am still having a hard time grasping the difference between induction and abduction as presented by Timmermans and Tavory 2012. Reflecting on my own undergraduate research, I feel as though what I was taught as being inductive research is much closer to the abductive reasoning described by the authors--namely, the iterative process of refitting data with new theories. I think my undergrad advisor would strongly disagree with the statement that an inductive, grounded theory approach does not generate new theories. In my undergrad methods classes I learned that theory generation is often suited by inductive, qualitative research, whereas theory testing is best suited by deductive, quantitative research. Understanding that research is and never has been quite so cut and dry, how should I come to think about abduction in research both, both qualitative and quantitative?

[donatellafelice]

when discussing the abductive research cycle, the authors discuss how innovation arises through tensions and collaboration, but in the conclusion they say their approach requires extensive theoretical and methodological training. It seems to me that a single person having a very in-depth theoretical knowledge would lead to less creative ideas, so wouldn't then the goal be not more methodological training but( given a relatively deep training) more interdisciplinary work? Or are they suggesting that most great research in this field will need to be performed by a team, each with specific specialties? Should we be focusing on how extensive a 'lone researcher's' knowledge is or placing the lone researcher with collaborators who are from different theoretical and methodological backgrounds?

[runlinw0525]

This is an very interesting article! Reflecting on the content, I'm curious: How does the idea of abductive analysis, as proposed by Peirce and discussed in the context of qualitative research, challenge the traditional grounded theory approach, particularly when it comes to embracing the researcher's unique position and theoretical background in the process of developing new theories?

[Brian-W00]

Based on the transition from grounded theory to abductive analysis pointed out in this paper, what could be the possible challenges researchers could encounter when adopting analysis, to be more specific, what could be the potential limitations in terms of balancing the integration of pre-existing theoretical frameworks?

[volt-1]

In the context of actively seeking negative cases or alternative explanations of Social Science research, what strategies can be employed to distinguish between genuinely contradictory evidence that challenges the emerging hypotheses and anomalous data that might be outliers rather than a theoretical revision?

[beilrz]

I am weak on the theory side of social science research, so I found this paper a little difficult to understand. For abductive reasoning, doesn't it require some already generalized insights or assumption of the system, how do I we engage abductive reasoning when we are facing a new area, and how do we know if it is appropriate to utilize such assumption? In the past, I usually see the previous theory as a guidance to ask an interesting question, rather than something to based my research on.

[Carolineyx]

Besides emphasizing "surprise" and "innovation", I haven't found abduction is so different from ground theory, would you mind explain further that how it is so special? Additionally, I am concerned that if we focus too much on the 'surprise' factor, it may divert us from the 'objective' of the analysis, thereby preventing us from presenting the whole picture of the data.

[h-karyn]

Although I have limited knowledge of sociology, I think the author's arguments supporting an Abduction Approach are valid in the greater field of social and behavioral science.

However, this paper reminds me of the debate of Theory Crisis in Psychology, especially in Social Psychology, which is now dominated by empirical quantitative investigations. Key aspects of the theory crisis include:

Lack of an Overarching Goal: many psychologists pay too much attention to investigating random phenomena instead of collaboratively solving a big problem.

Lack of Robust Theories: There's a concern that psychology lacks robust, well-developed theories that can reliably predict behavior and mental processes. This leads to a situation where empirical findings are abundant but not adequately grounded in a solid theoretical framework.

Over-reliance on Empirical Findings: Psychology, in recent times, has heavily emphasized empirical research, sometimes at the expense of developing and refining theory. This has led to a large number of studies that might be statistically significant but are theoretically disconnected or shallow.

Fragmentation of Knowledge: Due to the lack of unifying theories, psychological knowledge is often fragmented. Different areas of psychology might develop their own theories and models that are not integrated with each other, leading to a disjointed understanding of the mind and behavior.

This paper is concerning to me, because it seems like the Abduction Approach is behind this Theory Crisis in Psychology. I am curious, does sociology face the same challenge?

[HamsterradYC]

In article present the idea of "theoretical diversity" with the statement, "Knowing the theory means, in essence, knowing the theories." How would researchers to manage and in understanding and integrating perspectives from different disciplines and theoretical frameworks.

[naivetoad]

Reading the role of existing theories in abductive analysis, how might a researcher’s personal theoretical biases impact the abductive process? Is there a risk of confirmation bias, and how can it be mitigated?

[ddlxdd]

Considering the parallels drawn with Becker's 'Art Worlds', Merton’s patterns of serendipity, and Glaser and Strauss’s approach to teaching grounded theory, to what extent do the social interactions within a network of researchers and the routine practices involved in theory construction play a crucial role?

[JessicaCaishanghai]

In the article, it discusses the relationship between grounded theory and abductive analysis, the transformation of research strategy. How does the shift from a strictly inductive approach in grounded theory to a more abductive framework, as proposed in the passage, contribute to the construction of novel theories in the social sciences?

[michplunkett]

Like @beilrz, my social science research theory knowledge is not my strongest point, so I'll likely have to view it through the lenses I know. When it came to biological research, much like other research areas, I was always told that your hypothesis should stay fairly static as you get deeper into your research, sans something absolutely novel coming into play. When it comes to corpus analysis, adjusting your research goals/ends as your do your exploratory analysis feels intuitively like the right move.

On the topic of defamiliarization, could this be accomplished by the use of a NLP technician/researcher? Presumably the lead social science researcher is knowledgeable on the domain the are looking at, but a non-social science NLP researcher might be able to give them a purely analytical eye that lacks the social context for their research. This would, in theory, give the project a technically savvy eye that could look for patterns but wouldn't necessarily be limited by any prior assumptions on the data.

[yunfeiavawang]

The key differences between grounded theory and abductive analysis lie in their underlying principles and approaches to theory construction. Grounded theory emphasizes the inductive emergence of theories from empirical data, while abductive analysis focuses on the creative inferential process aimed at producing new hypotheses and theories based on surprising research evidence. Grounded theory aims to let theories emerge inductively from the data, while abductive analysis prioritizes the generation of new theoretical insights based on unexpected empirical findings. Given the philosophical definition, I'd like to know if there are any methodological guidelines for facilitating the process of theory exploration, and how we can confirm the robustness of the theory.

[floriatea]

Considering the emphasis on abductive reasoning in theory construction, how does the process of achieving theoretical saturation influence the depth and novelty of theoretical insights in qualitative research?

[icarlous]

The paper challenges the focus on causal influence and deductive reasoning in social science, advocating for the use of inductive and abductive reasoning. It questions how we can conduct research without predefined theories, noting that even field notes and coding are theoretically informed. This raises concerns about distinguishing such research from anthropological studies and underscores the importance of meticulous coding in abductive analysis.