Believability and Mathematical Reasoning
This blogpost was written by Lara Alcock, a professor in the Department of Mathematics Education at Loughborough University. Click on the link at the bottom of this blogpost to read more about Laraβs work, and the work of her collaborators, Dr Ben Davies and Prof Matthew Inglis. Edited by Dr Bethany Woollacott.
In this blogpost, Lara links logical reasoning in mathematics to psychological research on reasoning about everyday concepts.
Introduction
How believable would you say this is?
If π₯ is less than 2, then π₯ is less than 5.
How about this?
If π πππ₯ is greater than 0, then πππ π₯ is less than 1.
Both are true but, even if you know all the relevant mathematics, you almost certainly found the second less believable. How might that affect your reasoning?
Conditional Inference
I came to this question because I am interested in conditional inference, that is, in inferences from statements of the form ‘if π΄ then π΅’. Conditional inferences take four standard forms, as shown below. For which ones would you say that the conclusion follows in a logically valid way from the premises?
If π₯ is less than 2, then π₯ is less than 5. π₯ is less than 2. |
If π₯ is less than 2, then π₯ is less than 5. π₯ is less than 5. |
If π₯ is less than 2, then π₯ is less than 5. π₯Β is not less than 2. |
If π₯ is less than 2, then π₯ is less than 5. π₯Β is not less than 5. |
In mathematical logic, modus ponens (MP, top left) and modus tollens (MT, bottom right) inferences are valid, i.e., mathematically correct. Denial of the antecedent (DA, bottom left) and affirmation of the consequent (AC, top right) inferences are invalid, i.e., mathematically incorrect1. Evaluating validity might seem fairly simple with this elementary mathematical content; but students are immersed in everyday reasoning, where things are not so clear-cut. Compare the inferences above with those below, which have the same forms.
If you mow the lawn, then I will give you Β£5. You mow the lawn. |
If you mow the lawn, then I will give you Β£5. I give you Β£5. |
If you mow the lawn, then I will give you Β£5. You do not mow the lawn. |
If you mow the lawn, then I will give you Β£5. I do not give you Β£5. |
For this everyday content, you might feel inclined to accept the invalid inferences (top right, bottom left)2, and you would be reasonable in doing so. The conditional ‘If you mow the lawn, then I will give you Β£5’ would usually be interpreted as the biconditional ‘I will give you Β£5 if and only if you mow the lawn’. This reflects flexibility in everyday uses of βifβ: when interpreting everyday βif-thenβ sentences, people take account of context and conceptual content. Mathematics, in contrast, demands a single consistent interpretation3. This means that mathematical logic must be learned, and studentsβ reasoning will likely be influenced by experience with everyday language4.
Fortunately, everyday interpretations are well studied: cognitive psychologists have conducted extensive research on conditional inference 1,5. One thing that caught my attention is that people accept more inferences from conditionals that are more believable6. I wondered whether believability would also affect conditional inference in mathematics.
Could believability affect conditional inference in mathematics?
To find out, I needed mathematical conditionals that varied in believability, so my colleague Ben Davies and I set up a comparative judgement study7. This involved asking mathematics education researchers to generate mathematical conditionals that they thought would vary in believability8. We then asked these researchers, along with some mathematics undergraduates, to take part in an online study in which they were shown pairs of mathematical conditionals and asked which of each pair they thought more believable9. From their collective judgements, we used standard comparative judgement analyses to generate a believability score for each conditional.
We found that the researchers and undergraduates broadly agreed about believability, and that β as would be expected β true conditionals were on average judged more believable. However, truth and believability did not perfectly align. For instance, both researchers and undergraduates judged these true conditionals highly believable:
If π₯ is less than 2, then π₯ is less than 5.
If π is a multiple of 6, then π is a multiple of 3.
But some false conditionals were also judged relatively believable:
If π₯ is an integer, then π₯2 > π₯.
If line πΏ is tangent to curve πΆ, then πΏ intersects πΆ at only one point.
Indeed, these two false conditionals were judged more believable than some true conditionals:
If quadrilateral π has a reflex angle, then it will tesselate.
If π₯ is greater than 0, then πππ π₯ is less than 1.
This was helpful for our research plans, because it meant that we could design a conditional inference task using true conditionals with substantially varied believability. With our colleague Matthew Inglis, we are now studying which inferences mathematics undergraduates accept as valid from relatively believable and relatively unbelievable conditionals. Early results indicate that there is a mathematical believability effect, and that there are interesting and educationally relevant individual differences in studentsβ responses. These will be the subject of a future blog.
Acknowledgements
This work was supported by Leverhulme Trust Research Fellowship RF-2022-155 entitled βDoes Mathematics Develop Logical Reasoning?β.
References
[1] Evans, J.St.B.T. & Over, D.E. (2004). If. Oxford University Press.
[2] Cummins, D.D., Lubart, T., Alksnis, O., & Rist, R. (1991). Conditional reasoning and causation. Memory & Cognition, 19, 274β282.
[3] Alcock, L. (2013). How to study for a mathematics degree. Oxford University Press.
[4] Epp, S. (2003). The role of logic in teaching proof. American Mathematical Monthly, 110, 886β 899.
[5] Oaksford, M. & Chater, N. (2020). New paradigms in the psychology of reasoning. Annual Review of Psychology, 71, 305β330.
[6] Evans, J.St.B.T., Handley, S.J., Neilens, H., & Over, D. (2010). The influence of cognitive ability and instructional set on causal conditional inference. Quarterly Journal of Experimental Psychology, 63, 892β909.
[7] Jones, I. & Davies, B. (2023). Comparative judgement in education research. International Journal of Research & Method in Education, https://doi.org/10.1080/1743727X.2023.2242273.
[8] Alcock, L. & Davies, B. (2024). Believability in mathematical conditionals: Generating items for a conditional inference task. In S. Cook, B. Katz & D Moore-Russo (Eds.), Proceedings of the 25th Annual Conference on Research in Undergraduate Mathematics Education, pp.360-368. SIGMAA on RUME. http://sigmaa.maa.org/rume/RUME26_Proceedings2024-letter.pdf
[9] Alcock, L. & Davies, B. (2024). Believability in mathematical conditionals: A comparative judgement study. In Proceedings Fifth Conference of the International Network for Didactic Research in University Mathematics, pp.721-730. INDRUM. https://indrum2024.sciencesconf.org/data/pages/Proceedings_INDRUM2024.pdf
Centre for Mathematical Cognition
We write mostly about mathematics education, numerical cognition and general academic life. Our centreβs research is wide-ranging, so there is something for everyone: teachers, researchers and general interest. This blog is managed by Dr Bethany Woollacott, a research associate at the CMC, who edits and typesets all posts. Please email b.woollacott@lboro.ac.uk if you have any feedback or if you would like information about being a guest contributor. We hope you enjoy our blog!