A growing body of research indicates that asking questions about concepts students have not yet studied increases learning. In this episode, Steven Pan and Judith Schweppe join us to discuss how generative AI can facilitate this technique by generating prequestions.
Steven is an Assistant Professor of Psychology and the Director of the Learning Sciences Lab at the National University of Singapore whose research focuses on evidence-based teaching approaches. Judith is a Professor of Educational Science with a focus on teaching with digital media at the University of Erfurt. Her research focuses on working memory, long-term memory, and learning.
Show Notes
- Pan, S. C., Schweppe, J., Teo, A. Z., Indrajaya, A., & Wenzel, N. (2025). Using ChatGPT-generated prequestions to improve memory and text comprehension. Journal of Applied Research in Memory and Cognition.
- Josh Samani and Steven Pan (2022). Interleaved Practice. Tea for Teaching podcast. Episode 222. January 12.
- Samani, J., & Pan, S. C. (2021). Interleaved practice enhances memory and problem-solving ability in undergraduate physics. NPJ science of learning, 6(1), 1-11.
Transcript
John: A growing body of research indicates that asking questions about concepts students have not yet studied increases learning. In this episode, we discuss how generative AI can facilitate this technique by generating prequestions.
[MUSIC]
John: Thanks for joining us for Tea for Teaching, an informal discussion of innovative and effective practices in teaching and learning.
Rebecca: This podcast series is hosted by John Kane, an economist…
John: …and Rebecca Mushtare, a graphic designer…
Rebecca: …and features guests doing important research and advocacy work to make higher education more inclusive and supportive of all learners.
[MUSIC]
Rebecca: Our guests today are Steven Pan and Judith Schweppe. Steven is an Assistant Professor of Psychology and the Director of the Learning Sciences Lab at the National University of Singapore whose research focuses on evidence-based teaching approaches. Judith is a Professor of Educational Science with a focus on teaching with digital media at the University of Erfurt. Her research focuses on working memory, long-term memory, and learning. Welcome, Judith, and welcome back, Steven.
Steven: Thank you.
Judith: Thank you, thank you for having us,
John: And thank you for agreeing to come on at times that are a little bit challenging. Could you mention for our listeners what time it is where you are?
Steven: I’ll go first. It’s a bit after 4:30 in the morning [LAUGHTER] here in Singapore.
Rebecca: Starting your day right.
Judith: And for me it’s 10:30 pm, so I’m about to go to bed after this recording.
Rebecca: Best way to end the day. [LAUGHTER]
John: Today’s teas are:… Judith, are you drinking any tea?
Judith: Yeah, in fact, I’ve prepared a ginger and lemon tea.
Rebecca: A good one for the evening.
Judith: Yeah, exactly, that was the idea.
Steven: Well, I sometimes drink tea. Actually, I should drink more right now. I just have water.
John: And Rebecca,
Rebecca: I have Golden Monkey today.
John: That’s been a while.
Rebecca: Yeah, it’s been a while.
John: And I have an Irish breakfast tea today.
Rebecca: Awesome. So we invited you both here today to discuss an article in the Journal of Applied Research in Memory and Cognition that you both co-authored with three additional co-authors, entitled “Using ChatGPT Generated Pre-Questions to Improve Memory and Text Comprehension.” But before we discuss your study, it might be helpful to talk a little bit about what is meant by pre-questioning, and what past research has suggested about the impact.
Steven: Well, pre-questioning is a learning technique that involves attempting to guess the answers to questions about information you haven’t formally studied or know much about, and then there is, after you’ve attempted to answer those questions, the opportunity to learn the correct answers from some form of instruction, such as a lecture, watching a video, or, for example, reading a text passage. So you guess first, then you discover the answers afterwards. That’s pre-questioning. It’s similar to, it could you even be considered a part of a class of strategies that sort of involve what you might call pre-instruction learning activities: do something in advance of formal instruction, and pre-questioning, along with those other strategies or techniques, can improve learning in a variety of circumstances, improving, as I mentioned, learning from videos, learning from lectures, learning from text. There are a growing number of studies showing that it has these kinds of benefits.
Judith: What’s remarkable about this is that learners typically don’t know the answers to these questions, because they only learn afterwards, so they guess wrongly or can’t come up with an answer, they don’t get any feedback, but still they benefit on the criteria test questions after learning.
John: What are some of the possible explanations for why pre-questioning is so effective in enhancing student learning?
Steven: So this is an active area of research, and there are many possible explanations. Just to name some of them, for example, one explanation is that pre-questioning sort of activates attentional processes that either make you aware of your gaps in knowledge and want to focus on finding the answer, or when the answer is presented, you notice it better, or simply when the instructional opportunity happens afterwards, you are more focused, and you are more primed to learn from it. So that’s a sort of an attentional class of explanations. Then there is, for instance, an error signal type of explanation, where you make a guess, and usually it is incorrect because of the lack of prior knowledge or instruction, and your awareness of that error, particularly when the correct answer is presented, there’s a discrepancy. You guessed this, and then the answer is something else. And there are potential mechanisms in the brain that are responsible for learning from those kinds of situations, and the pre-questioning triggers those mechanisms. So that’s a second possibility. There are other ones that involve, for example, your memory of making the error. Okay, I guessed this, and now this is the correct answer, and you sort of integrate this, and that becomes a more powerful learning event, because it’s not just you learn something, it’s you made a guess first, it was usually wrong, and then you have the correct answer, and that becomes a richer, more integrated representation that helps you better remember that information. So, those are three candidate explanations for how pre-questioning can improve learning. There are more beyond that, but those are some of the ones that come to mind.
John: One of the other ones that was mentioned in your paper was that it also activates prior knowledge to help prepare you to integrate new knowledge into that framework.
Steven: Yes, that’s also a potentially useful way to consider how pre-questioning may enhance learning, and this idea that it may activate that prior knowledge, so you are then better able to integrate the new information, you may be primed to learn that information, and so on.
Judith: And that explanation might even be more important in real life in class, because even when you ask students questions for a new topic, they will have some prior knowledge that they can activate, whereas in the lab we usually use texts that they know hardly anything of, so that the importance of prior knowledge is even stronger, maybe in class.
John: And however it works, it does seem to work. [LAUGHTER] So it’s an effective technique, whatever the specific mechanism.
Rebecca: And despite the evidence on the value of pre-questioning, you know that the practice is rarely used in the classroom. So why might this teaching strategy be so rarely used?
Judith: I think there can be several reasons. For one, the research field is still very young, so there was not that much time for the evidence to distribute to the public, as compared to retrieval practice, for instance, which has been far better known in the past years. And another reason may be that it may also feel weird for a teacher, for an instructor, to ask their students questions that they know they won’t be able to answer, and that might refrain some people from using pre-questioning.
Steven: Yeah, those are, I think, very good reasons for why pre-questioning may not be used that widely. I’ll add another one that’s relevant to the paper that we’re going to talk about, which is this idea that you need the materials to engage in pre-questioning, and particularly for students who know nothing or very little about the topic, which is sort of definitional requirement of pre-questioning, it would be very difficult for them to come up with questions on that topic to practice with, and also even getting them from instructors, instructors commonly, for example, in surveys, they’ll say that they’ll provide practice questions on occasion, but typically after instruction has occurred, rather than beforehand. So all of this points as sort of a materials or implementation challenge. Not only do learners and instructors not necessarily know much about pre-questioning, and it also does feel weird to use it, but even if they wanted to use it, they might not have the materials, either the time to come up with the questions or access to questions that are appropriate for pre-questioning. So there are a host of obstacles that make it difficult, at least in many circumstances, to use this particular technique.
Judith: That was the main starting point for our study, because it is not only recommended as a teaching approach, but also as a learning strategy. But for learners to use it on their own, it’s a lot harder than retrieval practice, because, as Steven said, you cannot come up with new questions before you know about the topic, you’re totally reliant on somebody else providing these questions.
Steven: Yeah, exactly. And with retrieval practice, sometimes it can be as simple as: “take out a sheet of paper and write down as much as you can remember about whatever you’re trying to learn.” You just retrieve it. You don’t even necessarily need to have a database of questions to practice, but with pre-questioning you really do need those questions, and if you don’t have those easily available, then it’s very difficult to engage in the strategy or technique.
John: Which brings us to your paper. What are the primary questions you’re investigating in the study?
Steven: Well, there are sort of a sequence, we sort of went from very simple and basic, and we got more and more specific. The first question was simply, if you use generative AI to come up with these questions, and you use them for pre-questioning, do they actually work? That was sort of the most simple question, and that was the first experiment focus. We then advanced to, okay, how do those questions generated by AI compare to questions that a relatively expert human being creates? Are they similar or different in efficacy if you use both human-generated and AI-generated questions for pre-questioning? And that was the second experiment. And then, for the remaining two experiments, we went into more fine-grained issues, specifically comparing AI-generated questions made using what we called a sophisticated or elaborate prompt that was very carefully engineered to generate the questions we wanted to see, versus a simplified or more basic prompt that any person could use without much training, so different prompts to generate these questions, and we compared that to a condition that didn’t use pre-questioning at all, but did complete an alternative activity. In this case, it was studying an AI-generated outline, so that was sort of a more rigorous test of the effects of pre-questioning to see whether using pre-questioning compares favorably to spending the same amount of time studying something else.
Judith: It’s an additional question concerned whether answering pre-questions generated by either an AI or human would generate to a different source, because if you think of a regular learning situation, you might generate as a learner your question with an AI. Well, meanwhile, the instructor might also generate the exam questions with an AI, but they might also be generated by the teachers. So, it might be worth knowing whether there is any decline in the transfer from one source to the other, and the same with the more elaborate prompt, compared to a basic prompt. If maybe a student would use a more basic prompt than a teacher.
Steven: Right, and that issue speaks to this idea of, shall we say, transfer or generalizability. So you ask someone in pre-questioning a particular question, and then you assess learning using that same exact question, that’s one way to measure learning, but as Judith just said, it could be different questions that the instructor generates, either by themselves or using an AI, or using different prompts, or whatever, and so a critical question really is, “What happens when the assessment, the questions used on that assessment differ in some way from the questions used earlier for pre-questions?”
Rebecca: So we’ve talked a little bit about the research questions. Can you talk about the design of the study itself?
Steven: So each experiment, these were online studies with online participants, adult learners, and in each experiment the learners would be assigned to either a pre-questioning condition of some type, or in some experiments a control condition where they didn’t have pre-questioning, or they studied an outline, or some other activity happened. So that happens at the beginning, they’re randomly assigned to these different conditions, and then they would engage in some learning activity, for example, pre-questioning, attempt at least two pre-questions one at a time. All participants would then, and to be clear, all of these experiments involve learning from an encyclopedic text, in this case, I think it was about brakes, the function of brakes, different types of brakes, so it’s sort of a great proxy for certain kinds of information that you would find, for example, in an engineering course, something like that. And so they engage in the initial learning activity, such as pre-questioning. Then everyone reads this passage on brakes, which contains the answers to the pre-questions, and then after that, after a short delay, then they take an assessment of learning, which is a series of questions about brakes, which include questions the same as used for pre-questioning, but also other questions that were not necessarily exactly the same. That’s the broad design. Of course, across experiments, there are variations, so to get more specific, experiment one, the one condition was pre-questioning, where they attempted AI-generated pre-questions before reading the passage and taking a test. The other condition had no pre-questioning, so they just read the passage and took the test. In experiment two, it was AI-generated pre-questioning versus human generated pre-questioning. So one group got the AI-generated questions, one group got the human generated questions. They all read the passage, then they took the test. And then experiments three and four were either what we call elaborate generated pre-questioning, the prompt that was used, or basic generated pre-questioning, again referring to the prompt that was used, or studying an AI-generated outline. So, depending on the group, they would have different activities before reading the passage, and then taking the test.
Judith: And in any of the experiments in which we contrasted different questions, both types of questions were also used on the criteria test, so that we could see differences between the tested questions and the untested questions, and in some cases even between new questions generated by the same source, either both AI-generated or both human-generated, and transferred to new questions generated by the other source.
John: A very well designed experiment, and one of the things that I appreciated about this was that it was an experiment that was pre-registered. That’s not something that happens in all cases, in all disciplines. Could you tell us a little bit about this process? It’s something I think that came up because of people attempting to replicate some past studies and not always finding that results were replicable. So, could you just give our listeners a little bit of background on pre-registration of studies?
Judith: We try to do it in a brief manner. You’re right, this originated in what we call the replication crisis in psychology and social sciences overall, in which several well-published studies did not replicate, and so there’s a counter movement right now, the open science movement, and the idea is to restore trust in science and psychological research. We want to be as transparent, as open, about what we are doing as possible, and there are several reasons for why some research findings are not that reliable. There’s a tendency for significant findings to be published more often than non-significant findings, so that we are rewarded for getting significant results, and even more for getting our predictions right. There’s a study that indicates that more than 90% of hypotheses in psychological studies turn out to be right, and that’s not because we’re so clever, that might be due to some questionable research practices. So pre-registration is one way to make sure that we write down in advance what we predict, how we conduct our study, how we analyze our data, how we decide whether a certain data point is an outlier, how we decide whether to include any covariate. So to write that down in advance, so that we’re not subject to our own biases, so that may be the one analysis that brings about the best result that we’re convinced that that was the one we wanted to do right from the start. No, we can look that up because we’ve written that down, and to make it transparent to everyone, you pre-register publicly, so you are the one who decides whether others can see it, but when you submit a paper, you can include this pre-registration, and the reviewers can see that the date of the pre-registration is really before you collected your data.
John: You mentioned that so many studies, and this is true, it’s not just in psychology, it’s also true in pretty much all social science disciplines. And two of the explanations for that are publication bias and something called p-hacking. Could you just explain a little bit of those two things for our listeners?
Judith: Yeah, the publication bias, so that indicates, as I just said briefly, that of the data that are published, most of them are significant, and that’s not realistic, because a lot of studies we conduct, they yield non-significant results, and the idea is that this is due to several biases in the publication processes. So reviewers and peer-reviewed journals are more enthusiastic for significant findings, researchers don’t have that much time to write up their findings, so they tend to focus on those that are more publishable and leave the others in their drawers. So this is also called the file drawer problem, that there are a lot of non-significant findings in the file drawers of lots of researchers in all the ivory towers, and this leads to us overestimating the size of published effects, because we don’t know about this heap of unpublished studies, and that’s a problem for evaluating how important certain effects are for us. And p-hacking is one way where one type of questionable research practice, because p is the value that indicates statistical significance, and the field has agreed that p that is lower than .05 is the threshold for importance, and this statistical significance is often mistaken for importance. And there are several ways to analyze your data that can influence whether your findings yield a p value below this threshold, and all these ways to treat your data are called p-hacking. So, you can decide to include gender as a variable that was never part of your hypothesis, but you can always see whether it makes a difference what gender the participants are, and that might bind some variance, so that your main effect is significant. You can first analyze your data, and the p-value is just above .05 and you think, well, just a few more participants, and they might be below the threshold, or you might notice that there’s this weird participant, which must be an outlier, and that often happens after your first analysis, and helps you get below this very important threshold. This often is not outright fraud, it’s just us being human beings and being subject to our biases, and that’s why the open science movement thinks research science needs the self-control mechanisms to be open with our data, with our materials, and with what we predict. And we are both committed to that, so it was important for us to pre-register the study and also to publish the data and the material.
John: Another way of expressing the issue of p-hacking is that if you torture the data enough, it eventually will confess. [LAUGHTER]
Judith: That’s nice.
Rebecca: So gruesome. [LAUGHTER]
Judith: Something else that’s related to this is also HARKing, that’s hypothesizing after the results are known, so that you don’t actually make predictions, but rather post-dictions, and then there’s this analogy to the Texas sharpshooter who draws the target after his shots, so that that is easy to hit bull’s eye, and that’s also often used to compare this research.
John: But your results are not subject to any of those issues, because you’ve done all these things to prevent that with pre-registering and sharing the data.
Rebecca: Seems like a good transition to results, huh? [LAUGHTER] Can you talk a little bit about the impact of providing the AI-generated pre-questions? I imagine that folks are really interested in this particular piece because it might save time.
Steven: So we were very excited to see that the AI-generated pre-questions were effective at enhancing learning, or producing meaningful learning, shall we say, across all the experiments. So, for example, in experiment one, just presenting two, I think, AI-generated pre-questions before reading a text passage, that was enough to enhance learning from that passage, and that enhancement extended not only to later questions that were exactly the same as the questions presented in the beginning, but also new, never before seen, questions, so there was a generalizable benefit to some extent. In the second experiment, we found that the AI-generated pre-questions yielded similar levels of learning as questions generated by competent human beings, so that suggests that you don’t necessarily need a human being to invest the time and effort to come up with the questions. An AI can potentially, an AI chatbot can come up with similar questions of comparable efficacy, and then in the third and fourth experiments, again, they were similar designs, those two experiments, both AI-generated pre-questions that were generated using a more sophisticated, detailed prompt, as well as those generated using a very simple prompt, they both enhanced learning relative to studying an AI-generated outline, so that shows the power of pre-questioning extends to not only using AI to generate the pre-questions, but even compared to using AI to generate some alternative materials that are used for some alternative activity. So those are the broad outline of the key results. Now there’s, of course, that other finding about, or other issue about what happens when the questions on the test are not the same as those that you were practiced on, or they were generated by a different source, and whether, for example, you practice with AI-generated questions, but then some of the questions on the test are generated by a human being or generated by a different AI prompt, or things like that, and I think one general way to describe those results, because they varied across experiments, was that we could still see across experiments some evidence that learning still benefits in those cases, so that it’s not essential that the questions that are used on assessment have to be exactly the same as the ones that were seen earlier. There is some degree of generalizability, although there was sort of a trend, I would say that as the questions get more and more different in source for the content targeted, the learning benefits may not be necessarily as strong, at least numerically, but there’s still evidence that the learning is not restricted to the exact same questions that you saw earlier.
Judith: And we were very happy about that, because that’s not always the case in pre-questioning studies, because often these effects are restricted to post-test questions that target the very same information, so that the transfer to these untested questions really excited us about these findings, and also the case that this very, very basic prompt, which just said, “Can you generate two questions for this passage?” worked, that really sounded feasible.
Steven: One thing I would add to that, we were really very excited by that finding, because literally that means that anyone practically could just use a prompt that is one line and come up with questions that are useful for pre-questioning or potentially other purposes. One thing I did want to note, though, is that when we were conducting the sequence of experiments that we were using ChatGPT, the version of ChatGPT was getting more and more advanced as the experiments progressed. So, by the time we got to the latter experiments and we were comparing different types of prompts, we were using a more advanced GPT version, and in that case, the questions made using, again a very elaborate prompt versus a simple prompt, they seem to be equally effective for the most part, which is again I think in part a reflection of how advanced the technology has become, and also how just generally how easy it is now to use.
John: Given that you’ve got these results, and it doesn’t require a very elaborate, detailed prompt, or even much knowledge of the subject, might you suggest that this would be a useful way of students learning things more deeply as part of their studying? Because students often use it to generate an outline, which was one of your control conditions there, and that was not as effective as having these pre-questions. Might this study suggest that this would be an effective study technique for students to use on their own when they’re learning new things?
Judith: I guess we would say yes. That was the idea, yeah. That was the hope to find evidence that it can work if they generate the questions on their own. We might also add a cautionary note to that, because we found transfer to untested questions, but that may not necessarily be the case for other questions. We noticed that there was some overlap between the questions generated, because all of them were constructed in a way that they were integrative, so they targeted more than one piece of information, and that resulted in somewhat overlapping questions, so the suggestion is a bit different compared to other studies, in which the new questions are totally different, and it might still be more helpful to get the pre-questions from your instructor, who will also post the questions for the exam, because they also tell you what this person thinks is most important, and that may be a valuable addition, but if your instructor doesn’t want to give you any pre-questions, because they don’t believe it’s effective, or because they don’t want to waste their good exam questions on students preparing for learning, that might still be good alternative to ask a generative AI to be a question generator.
Rebecca: I’m imagining that faculty hearing this are excited about the possibility of integrating pre-questions into their teaching methodologies. Do you have some advice about how they might do this, how they generate effective questions, or even like how many questions would be useful?
Steven: So, there are some tentative recommendations that we could offer, but again, these are tentative because we have barely scratched the surface of implementing pre-questions. We are still sort of exploring the efficacy and mechanism stage. Nonetheless, in terms of possible ways that the pre-questions could be generated in an effective manner, well, firstly, one thing I would note was you mentioned number of questions, so it seems at least in the studies that have been conducted thus far, the number of questions is not that many. For example, I don’t think we went more than four or so pre-questions, depending on experiments, and sometimes there’s just two. Why is that? Well, in part, because if you ask people many, many, many questions on something they know very little about, and then they have to try to find the answers afterwards, trying to keep in mind, say, I don’t know, 10, 20 questions is difficult, and so a limited number of targeted pre-questions on perhaps the topics that are most crucial is probably a good way to go in terms of number. Now, in terms of type of questions, you just mentioned the questions that we used were integrative questions, and one of the reasons why, besides what was mentioned, is that there’s some prior work on different types of pre-questions. Some pre-questions that have been investigated are so-called detail-oriented, that might just target a little fact or a number or a little detail, and then there are these integrative questions that might connect two concepts or different parts of a text passage, what have you. Those seem to be more effective at improving learning. So, in addition to the number of pre-questions, the type of pre-question probably I would recommend to instructors to think more about integrative pre-questions. Even if you want a student to learn one particular small detail, maybe couching it within an integrated pre-question might actually be better. And then there is a third sort of practical point that I just wanted to raise, which is that definitely keeping in mind that pre-questions, in a way, serve as almost like a introduction to the topic, so these are not the kind of questions used necessarily to really assess mastery in a retrieval practice sense: “Explain this concept or this theory that you’ve just learned.” If you ask such a question to someone who has not learned that concept, that theory, that question may be so difficult they don’t even know how to begin to answer such a question. Then it’s probably not the best type of question… more like a question that is in some sort of currently perhaps nebulous sweet spot of they can come up with some sort of guess, it’s probably not correct, but it’s also not so impossibly foreign that they look at this question, and think, “Well, I don’t know where to begin.” So, those are at least three tentative recommendations that I have about how pre-questions could be developed by instructors for effective use.
Judith: Something else I would like to add is that you can also first think about what do you want your students to learn the most. So, what are the most important aspects? Knowing that these pre-questions work best for what they ask for, you could make use of that and target the most important aspects with the pre-question. That’s also something related to our study, to think of the cognitive levels associated with your learning objectives, and to see that the questions are generated in a similar way, so that also guides the learners. We also looked at this in our pre-questions, and noticed that for the first experiments, the AI-generated questions were mostly on the lower cognitive levels of Bloom’s taxonomy, so mostly remember and understand, whereas the human-generated questions we used that were used in an earlier study, use comparisons, and that was not the case with our AI questions. So that was the focus of the fourth experiment, where we engineered a prompt that asked for comparison questions, and then the AI generated these questions, but only then. So that might also be something to take from this study that the Gen AI might need an additional prompt to generate this kind of question, but that might also be the case for a human being that thinks of questions that maybe we also need an additional prompt to not just ask remember and understand questions, but also target these higher levels that we name in our learning objectives, and to kind of reverse engineer our questions from what we want learners to take away from the learning material.
Steven: Yeah, and if I may add that Judith raises a very important point, which actually qualifies the finding about basic versus more elaborate prompts, even if you use a basic prompt, you know, generate questions, you can still add in there, generate questions that involve comparison. You can add some details that’s not necessarily required multiple premises in a very complex prompt, but some of those things actually might be necessary through even a process of a little trial and error to come up ultimately with questions that meet certain characteristics that are desirable for learning, including higher order type questions, questions that tap different cognitive levels, that promote the types of learning objectives you want to promote.
John: One of the things you were doing in the study is comparing AI questions with human-generated questions. Was there any apparent differences between the types of questions that humans generated and the types of questions that AI generated?
Steven: We did have sort of as an additional measure in the study… we didn’t tell them these were AI-generated questions, we asked them to guess who made these questions, and we would just give them all the questions that we had in a particular set and some were human-generated, some were AI generated. We just asked them, and they just could not tell, and so this is, I think, part of a larger challenge now, which is differentiating between AI-generated and non-AI-generated materials. Now, not only are the questions as effective as those generated by human beings, but also human beings cannot tell that they were AI generated, but I thought that was kind of a fun little thing, and it speaks to this larger challenge that we have, whether it’s trying to detect AI use in our students’ essays, other things like that.
Rebecca: We always wrap up by asking, what’s next?
Steven: Well, in my research group, and also in collaboration with Judith, we’re really interested in exploring a variety of aspects of pre-questioning, in terms of, for example, ways to implement it in authentic educational circumstances more successfully. So, for example, studies in the classroom, or studies exploring how you might integrate pre-questions into learning activities that are very common, like reading a textbook chapter, things like that. So we’ve explored AI-generated pre-questions, we’re very interested in implementation that could be easily translated to real-world practice, so that is one area of very active interest related to pre-questioning.
Judith: And then a related question in our labs… Well, I thought of it because it’s somewhat related to the current idea, because I’m currently working on a study with a PhD student where we want to investigate whether spacing, distributed practice, which is recommended both as a teaching strategy, but also as a learning strategy, but it’s investigated almost always so that the entire schedule, sticking to the schedule, that’s decided by the experimenter, and our idea was we really don’t know whether that works as a learning strategy, so we implemented it in a way that gives some responsibility for the learning, for sticking to the schedule, to the learner, and see whether it’s still an effective strategy, and also to see whether something like gamification might help learners become better spacers.
John: You’re doing some great work here, and actually that reminds me a little bit of the last interview with Steven and his co-author on another study that looked at classroom implementations of interleaved practice, because a lot of the interleave practice studies had been done in the lab and it was nice to see what happens when it was actually introduced in the classroom environment. You’re both doing some really good research, and we’re curious to see what you’re going to find in those new studies. Thank you.
Judith: Thank you.
Steven: Thank you.
Rebecca: Yeah, thank you so much for joining us.
Judith: Yeah, thanks for the great conversation.
[MUSIC]
John: If you’ve enjoyed this podcast, please subscribe and leave a review on Apple Podcasts or your favorite podcast service. To continue the conversation, join us on our Tea for Teaching Facebook page.
Rebecca: You can find show notes, transcripts and other materials on teaforteaching.com. Music by Michael Gary Brewer.
[MUSIC]