7 July 2021

Who Teaches the Teachers? Teaching Science Thought and Practices

K. E. Saavik Ford                                                             Dennis M. Robbins
CUNY Borough of Manhattan Community College          CUNY Hunter College
American Museum of Natural History                              American Museum of Natural History

Most science educators tell us they would like to teach students how to think scientifically — they’d prefer students not learn a long list of facts or the details of some classification system, but instead, students could learn and engage in the practices and thought processes of working scientists. Such practices would include proportional, graphical, and analogical reasoning (reasoning by analogy). Students would reason about and with data, including hypothesis testing and experimental design, and control of variables thinking.

Yet, traditionally curricula do not target these specific learning outcomes; cookbook labs where the outcome is preordained, or problem sets with known endpoints are the norm. Instead, we expect students to pick up their creative scientific skills through some kind of osmosis. This is both inefficient and destined to amplify pre-existing social inequities — we put the most marginalized students at the biggest disadvantage, though this tradition doesn't do students from dominant groups any favors either.

We noted this mismatch and designed a course, Science Thought and Practices, to directly address these conceptual stumbling blocks with undergraduate physics majors early in their careers (including first-semester freshmen). The course was developed, and runs, as part of the AstroCom NYC program (funded by the National Science Foundation), which is designed to recruit and retain underrepresented minorities in astronomy and astrophysics. In that context, the course has met with spectacular success — each semester, students come out with improvements on measures of their scientific reasoning ability. They also often tell us it is the best or most useful science course they’ve taken. The course is inquiry-based so that it is both hands-on and ‘minds-on — we seek engagement and collaboration at all times. In the context of the AstroCom NYC program, the course serves as preparation for a 10-week summer research experience where students as early as rising sophomores can make meaningful contributions to ongoing astronomical research.

An Investigation

To give a sense of how an actual investigation works, we describe the first investigation — mass versus volume — below. We begin by handing learners a graduated cylinder and asking them to discuss with their neighbors what the instrument is, and what variable(s) it can measure. We then lead a discussion wherein our goals as educators are to elicit broad or universal student agreement that it is:

  • a graduated cylinder
  • that measures volumes of liquids
  • in units of mL
  • with a range of 0.0-10.0mL
  • to a precision of 0.1mL
  • that a simple description of volume is ‘the amount of space something takes up’

We repeat a similar exercise with a USB connectable electronic mass balance, and ask the students to consider an investigation that we could conduct using these two variables; in particular, what variable would they choose for the independent variable, and which would they choose for the dependent variable (which we define), and why? Which variables would they want to hold constant? We are careful to let many voices speak and to never say ‘yes’ or ‘right’, especially after the first volunteer (even though we may agree with them). It takes many weeks of this behavior (typically 5-6 weeks) for actual undergraduate students to learn and trust that you will always listen to what they actually think, rather than a more traditional seeking of ‘the right answer’ in order to move on. Eventually, we can generally get agreement that volume makes the best independent variable because it is easily manipulated visually using the graduated cylinder, and we can let mass be the dependent variable; we should also ensure the substance is held constant. We present them with a familiar liquid, water, on which to experiment. Next, we agree on the experimental procedure (collect data at 0.0, 1.0, 2.0, etc., up to 10.0mL, agree on what 1.0mL looks like among all groups), with a careful discussion of and attention to zeroing the mass balance (place the empty cylinder on the balance and zero it). We set up the data to be collected electronically in the LoggerPro (made by Vernier). This includes a graph with volume (the independent variable) on the horizontal axis and mass (the dependent variable) on the vertical axis, with appropriate axis limits, labels, and units.

Finally, we pause. We give each student a post-it of mini-graph paper and ask each one to predict graphically what their data will look like. Responses do vary, and we allow a period of Think-Pair-Share; the main goal is to get each student to give a reasoned case for their prediction.

The investigation is then conducted in pairs; once the data are collected, we introduce them to the analytic functions of the graphing software. We encourage learners to choose a proportional fit (rather than a linear fit) and express the criteria for choosing it (data go through (0,0), and are line-like). The software provides the equation in terms of the variables (m=AV), the value of the best-fit slope (A), and the RMSE (root mean squared error) of the fit. We give them a form of words to discuss the slope: ‘On average, the mass increases by 1.0 grams for every milliliter of water.’ This gives the slope a physical meaning and we will continue to use this throughout the course. We also note that the slope has its own units of g/mL.

Then we hand around a new substance, 90% isopropyl alcohol. We ask students for new graphical predictions on the same post-it, and their reasons for it. If they choose a smaller slope (compared to water), we want to hear that they think the alcohol ‘feels lighter’ so they think there is ‘less stuff’ in a given volume of alcohol (but if they draw a larger slope prediction, we want to hear that they think the alcohol ‘feels heavier’, etc.). They repeat the experiment and analysis and we compare the slopes.

Finally, we hand around a third substance: Karo Syrup. We ask for predictions in the same way and send them to conduct the experiment. However, Karo Syrup is a miserable experimental substance to work with: it sticks to droppers, the sides of the graduated cylinder, it comes out in great globs… you get the idea. We encourage students to do their best and stick with it. Their new data are generally fit by a line but have a much wider variance than either of the previous two experiments. We then introduce the concept of the RMSE, both visually, from the graph, and algebraically — we motivate and explain each part of the construct (‘error’ is Δyi; if we just sum those up, the negatives and positives cancel, so let’s square it; but we do want the mean over all the points; but now our answer is in different units than our measurement, let’s square root it). We then discuss the utility of models for making predictions, and how to consider or characterize the predictive power of a model (Lest someone panic, this is not the only thing we teach them! In later meetings, we explicitly introduce investigations where RMSE does poorly at characterizing models).

The Workshop

In order to build on the success of the course, we created a professional development curriculum (PD) to allow other educators to learn how to teach the same course, and/or adapt aspects of their own curricula to effectively address some of the same topics. We used an AAS-EPD mini-grant to offer a very intensive, weekend-long version of this PD curriculum to postdocs and faculty during the workshop weekend prior to the January 2019 AAS meeting.

We like to say that people teach as they were taught; some of the strategies we employ in teaching the course have more to do with creating effective interpersonal interactions at key moments of student openness, rather than any particular curriculum. Demonstrating those techniques on future teachers as learners highlights many of these effective interactions. The weekend PD course consisted of four lessons or investigations in which we taught the instructors as though they were our undergraduates, followed by a reflection and discussion period to identify key goals and effective practices. Each lesson was selected from a different sequential unit of the course: Proportional Relationships (using mass vs volume of three liquids); Inversely Proportional Relationships (using pressure vs volume of air); Linear Relationships (using pressure vs temperature [Celsius] of air); and Consolidation of Variables (an experimental derivation of the ideal gas law); a further unit, Univariate Analysis, was neglected in favor of allowing participants to prepare to team-teach an investigation of their choice.

Note that during the workshop we did not conduct the investigations with postdocs and faculty any differently than with first-year undergraduates — while senior people obviously already know the material it is rare that they’ve thought about it in this pedagogical format. Most senior people who have been exposed find the approach both familiar and refreshing — a way of bringing what we actually do in our research lives into teaching, and in doing so providing a way of equitably scaffolding student development towards becoming independent researchers.

After the weekend workshop, we offered a series of two-hour 'in brief' workshops (during the main AAS meeting) where new participants were exposed to an example investigation lesson taught by the weekend intensive students. While the weekend intensive students were teaching, Dennis or Saavik would ‘hover’ and provide guidance and suggestions (quietly) on how to direct or re-direct the lesson. In this way, weekend educators received both on-the-spot practice and practical feedback, and new educators got a taste of what a different kind of lesson could look like. Overall, the feedback was extremely positive, especially from the weekend intensive participants.

Unfortunately, the workshop is so condensed that it is extremely exhausting to run at the AAS (both for instructors and participants). We are considering other ways to disseminate the techniques, including week-long workshops, either in New York City or at other geographic centers where many learners can gather. If you would be interested in learning more, hosting us at your institution, or attending an extended workshop in NYC, please get in touch with Saavik at sford@amnh.org.