Understanding more about patterns of scientific inference. Part III – Abduction
Understanding more about patterns of scientific inference (Part III – Abduction) By Huayi Huang “As inquisitive learners, people like to try to explain things that surprise us. Abductive learning recognises that we do this through spontaneously generating explanatory ideas which come to hand, as we try to informally explain the surprising data or evidence” Whilst Deductive and Inductive inferences focus on movements, between specific and general ideas in learning, standard accounts of deduction and induction tend to draw examples from the formally published parts of what researchers’ actually learn. In particular, academic journal papers, funders’ reports, etc. tend to present highly sanitised versions, of the day to day learning from researchers’ professional lives and work. Treated as descriptive accounts of research-reality, such published accounts usually miss the various hunches, clues, metaphors, possible patterns, and informal explanations interacting with research practices. As inquisitive learners, people like to try to explain things that surprise us. Abductive learning recognises that we do this through spontaneously generating explanatory ideas which come to hand, as we try to informally explain the surprising data or evidence. This in essence seeks to turn the challenge (of the surprise!) to our currently held beliefs, into a more plausible thing – when the surprising data or evidence is viewed as a concrete example of one or more of the explanatory ideas we spontaneously generate. In relation to Deduction/Hypothetico-deductive inference, new ideas are not so much developed based on those already accepted as certain (Deductive) or most likely (hypothetico-deductive) by a knowledge community. Instead the focus of reasoning is on developing abstract ideas (and sharing accounts of them) so as to minimise surprise for the learner. We all spontaneously generate ideas to explain what we encounter, as we live. These are ‘informal’, in the sense that the ideas we explain by are not necessary already part of current published research literature. Some qualitative research experts link such Abductive inferences to the idea of qualitative research being the systematic empirical inquiry towards meaning (e.g. for both the researcher + researched). In another words, evolving our systems of thoughts and evidence in such a way as to make the new evidence (e.g. from primary data collection) meaningful – within the evolved (i.e. ‘more educated’ 🙂 system of thought. There are of course limits in practice however, in the extent to which the ideal of making all new evidence meaningful within the same system of thought, actually happens during 1 research project. In helping us to make sense of each others’ everyday social and human experiences, an academic qualitative research interest is to help readers discover and come to terms with the alternative or multiple meanings for any significant life event. Gilbert Harman’s characterization of abduction as “reasoning to the best explanation.” is a common experience of qualitative researchers who try to explain their data. Professor Robert Miller in his 2003 A-Z of Social Research (https://uk.sagepub.com/en-gb/eur/the-a-z-of-social-research/book211452), characterises this ongoing process of reasoning to the best explanation by analogy from the Darwinian idea of survival of the fittest, based on a foundation of reality being a singular thing accessible equally to all. The population in his case is not so much a biological one, but referring to the ‘population’ of one or more explanatory ideas spontaneously generated, as we learn about some data/evidence. As more interviews are done for example, some of a research team’s earlier explanatory ideas might be dismissed as ‘less fit’, in light of the current pool of qualitative research data (and therefore not followed up as a study progresses). In summary, Abductive inferences try to build explanatory patterns that turn a set of data or evidence, into one of least surprise/highest predictability to the knower. Abductive accounts of reasoning largely forego the historical issue of how particular explanatory schemas have come about, instead focusing on the utility of explanations in making our future predictable. As scientist go about their professional lives, abductive inferences (in the sense of hunches, clues, etc.) only become visible as one zooms down from a high level view, into the experiences recorded in specific project or personal journals, lab books, and so on. The messy world of near misses, inspirational moments, and the confusions set aside, are an all too often unpublicized part of the research world. Yet these are human experiences we all share. Due to the experiential and conceptual worlds typically accessed through qualitative research data, the world of abductive reasoning is all too familiar to the qualitative research experience. The human project of evolving our thinking towards a more consistently meaningful and therefore predictable world within our systems of thought (abduction), can progress without necessarily focusing on whether we are moving from smaller to bigger ideas (induction), or wider to narrower ideas (deduction).
Analysing Focus Group data
Focus Groups Focus groups are essentially a “series of discussions to obtain perceptions on a defined area of interest” (Krueger and Casey, 2009: 2). Typically, these seemingly informal discussions occur in a group of no more than 12 individuals (Asbury, 1995; Smithson, 2008), who share some common characteristics (Franz, 2011) and “have a direct experience of the topic” (Acocella, 2012: 1127), and, thus, are treated as ‘experts’ (Stewart and Shamdasani, 1990). These discussions are facilitated by a moderator. Analysing focus group data The analysis of focus group data can be quite complex, and involves both the verbal content of the debate and the non-verbal communication. Still, focus group data are often under-analysed and used, for example, to simply report emerging themes and coding frequency. When preparing this post, I did a quick online search for what people say about focus group analysis and immediately found comments such as the following: “analyzing focus group interviews is not greatly different from analyzing one-person interviews.” “I personally think that the analysis of focus group data is not much different from the analysis of any other kind of interview data. Most people who ask for specific analytic tools in focus groups do so because they don’t have experience with analyzing other kinds of qualitative data, so they think there must be something unique about focus groups.” … Needless to say, they couldn’t be more wrong. Focusing exclusively on individual participants’ accounts (which would be the case if “the analysis of focus group data [was] not much different from the analysis of any other kind of interview data”, as suggested above) would mean that we are overlooking the most unique and valuable aspects of focus group discussion – the group’s views and group dynamics. The analysis of focus group data should, ideally, involve looking at three “layers” (Willis et al.) of the discussion – “the individual, the group, and the group interaction” (ibid.). Importantly, these different layers are not necessarily “steps” – thus, in practice, the analysis of one layer is difficult to separate from another (although I think there are also some guidelines that suggest doing so). When you investigate the “individual” layer, for example, not only do you want to pay attention to all views and opinions expressed by one group member throughout the discussion, but also to how this participant “functioned” within the group (were these views relevant to what was being discussed, were they similar to/conflicting with the group’s views, etc.), to whether these views changed throughout the discussion (were they influenced by the discussion? Did they emerge because of some other participant’s views?), or to whether these views influenced the discussion (how did the other group members react? Were these views supported? Or ignored? Was the participant silenced?). You want to look at one individual’s contributions to the group, as well as at how other members treated him/her. You may also see if this person tried to reinforce his/her opinion more than once and, again, how the group reacted. Similarly, when you focus on the “group” and the “group interaction” layers, it is not possible to ignore the individual roles. When you look at whether the group achieved a “common ground”, or a kind of a shared identity or self-positioning, you need to pay attention to how the individuals constructed this identity together and how each participant responded to the construction of this identity (did he/she contribute to it, and self-positioned him/herself as a group member? Did he/she distance him/herself from the group?). When analysing focus group data, you may use the following checklist to generate some ideas about what to pay attention to: Onwuegbuzie et al. (2009) propose another useful structure for focus group analysis which is centred around a combination of different data analysis techniques – constant comparison method (which may involve comparing themes emerging in different focus groups), content analysis (e.g. coding the content of the discussion and paying attention to coding frequencies), and keywords-in-context and discourse analysis, both of which involve careful examination of the use of words, and language as a whole, throughout the focus group discussion (at both the group and individual levels). I would not even want to try to discuss focus group analysis in much detail in such a short blog post, but I do hope that the above gives you a general idea of what it may involve (and that you will understand that it is definitely NOT the same as individual interview analysis!). Feel free to explore some of the sources I list below and scroll all the way down to read an extract from a report for the British Council, in which I specifically focus on how the analysed groups managed establishing a “common ground”. REFERENCES Acocella, I. (2012). The focus groups in social research: advantages and disadvantages. Quality & Quantity. 46 (4), 1125-1136. Asbury, J.-E. (1995). Overview of Focus Group Research. Qualitative Health Research. 5 (4), 414-420. Franz, N.K. (2011). The unfocused focus group: benefit or bane? Qualitative Report, 16 (5), 1380-1388. Krueger, R., & Casey, M. (2009). Focus groups: A practical guide for applied research. 4th Ed. Thousand Oaks, CA: Sage Publications. Onwuegbuzie, A., Dickinson, W.B., Leech, N.L. & Zoran, A.G. (2009). A Qualitative Framework for Collecting and Analyzing Data in Focus Group Research. International Journal of Qualitative Methods, 8 (3), 1-21. Smithson, J. (2008). Focus groups. In Denzin, N. and Lincoln, Y. (eds.) Handbook of Qualitative Research. 2nd Ed. Thousand Oaks, CA: Sage Publications. Stewart, D.W. & Shamdasani, P.N. (1990). Focus Group. Theory and Practice. Newbury Park: Sage. Willis, K., Green, J., Daly, J., Williamson, L. & Bandyopadhyay, M. (2008). Perils and possibilities: achieving best evidence from focus groups in public health research. Australian and New Zealand Journal of Public Health, 33, 131-6. Extract from my focus group results section in Galloway, Kriukow and Numajiri (2017: 29) “A deeper analysis revealed that the discussed topics and opinions shared by the groups varied in relation to the ways the participants were establishing common ground and positioning the group as a whole, as well as self-positioning themselves within the group (Hydén and Bülow, 2003). For example, in the Hubei focus group, where all the participants majored in English education, they established a collective voice of ‘English majors’ quite effortlessly, and constructed their narratives as a group of ‘experts’ in their field. They spoke of
Tips on how to ‘unlock’ your data analysis
A common problem with analysing qualitative data, which both students and more experienced researchers face, is… simply getting stuck with it. It’s like writer’s block, but in data analysis – you may have all the necessary components in theory, but you are not sure what to do with them. You are not sure what to analyse and how, where to look, what to focus on, etc. Here are some steps that always help me “unlock” my data analysis and dig deeper into the data (the order is random, you may apply some, or all of them, and at various stages of your work): Read through the data, without worrying about coding. Try to get the overall “feel” of it, occasionally recording some thoughts that may occur (e.g. if you see patterns across the data, some similarities/differences, something that you think could turn out to be significant, etc.) Try to sketch a simple model of what you see in the data. You can either do it with a computer software (data analysis software, such as NVivo, or specific software for designing models – I have used https://www.draw.io/) or by hand. I quite often do it on a separate piece of paper, which results in hundreds of pages with weird and difficult to interpret “hieroglyphics” scattered around my house. In the model, try to “theorize” about your data, potential links, relationships and associations between different elements of your project (e.g. the themes you develop, or certain experiences/beliefs described by your participants), possible causal relationships (e.g. this seems to result in that) and other things you notice. This is supposed to be theoretical, so don’t be afraid to make assumptions and jump into conclusions – this is Exactly what you need. As a result, you have a whole bunch of possible relationships, assumptions and “hunches” to investigate and check if they are, in fact, true. Write about your data (you can do it as a free writing exercise) – again, as with the model, write about what you have found, as well as about what you think you may find, write about everything that seems to be relevant in any way. Similar to drawing the model, this may help you “open up” your data, come up with ideas that you may later explore and investigate (and this will be the actual analysis part of your data analysis) Run word frequency queries – to see which words appear most frequently in your data. This sometimes helps to draw your attention to issues that you previously did not pay attention to “Dig deep” into what the content of what the participants said meant to them – go beyond simple factual information that they convey, try to see who said something, how they said it, under what circumstances, when they decided to talk about it (e.g. which interview questions prompted them to give a certain response), see what followed these thoughts, what they chose to speak about (e.g. do they always talk about stressful experiences at school when you ask them about their childhood?), what kind of emotions they seem to experience when discussing particular topics, etc. Conduct a detailed “within-case analysis” – focus on each participant individually, considering some of the points made above, as well as whether one participant changed his/her opinion within the same interview, how he/she reacted to similar questions asked at different stages in the interview, etc. If you have a coding framework, analyse it in detail for one participant at the time. Try to draw a model for each participant. Write a case summary – summarize what a given participant said, what your thoughts and observations based on this are, etc. If you have more than one data set for an individual participant (e.g. an interview transcript, a diary, open-ended questionnaire responses, etc.), treat them as one “case” and explore all these data sets in detail. Conduct a detailed “cross-case analysis” – follow similar steps but this time with the aim of comparing the “cases” (e.g. different participants), looking for similarities and differences, comparing these similarities/differences in relation to (demographic) characteristics of the participants (e.g. did those with more teaching experience express more positive attitudes towards something than those with no experience, did men express similar opinions to women, did students share similar views to their teachers, etc.) If you have a coding framework – create matrices that compare the coding between the participants in relation to the above kind of characteristics (in NVivo, this is called a matrix coding query) Check whether the participants repeat similar points, whether there are re-occurring patterns, when these patterns occur Investigate “negative cases” – these are the cases (e.g. participants) that, for any reason, do not “fit” the rest of cases or your emerging “theory” or explanation (e.g. what one single participant said is completely the opposite to what all, or the vast majority of, the others said). I love investigating negative cases – remember that they don’t “ruin” your theory, or your analysis in general, but rather enrich it and provide a lot of new insights Do you have any other tips for “unlocking” your data? Please share them in the comments!