Computational Narrative Activities: Combining Computing, Context

Chapter 12

Computational Narrative Activities: Combining Computing, Context, and Communication To Teach Chemical Concepts Downloaded via UNIV OF ROCHESTER on May 15, 2019 at 18:19:35 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

Steven M. Singleton* Department of Chemistry, Coe College, 1220 First Ave. NE, Cedar Rapids, Iowa 53202, United States *E-mail:[email protected].

This chapter describes the author’s experience developing computational narrative activities (CNAs). These interactive activities utilize current computing technologies (python and Jupyter notebooks), and use established assignment design strategies to help students learn physical chemistry concepts while developing skills in problem solving, writing, and applied computing. Design considerations and implementation details of CNAs are described in the context of an undergraduate physical chemistry course.

Introduction Using computing as a tool for teaching and learning has history as old as the computer itself. A 1953 Journal of Chemical Education article by Scott (1) relates how Edward Teller and co-workers used MANIAC (Mathematical Analyzer, Numerical Integrator, and Computer) at Los Alamos National Laboratories to develop a unified model of liquids akin to the kinetic theory of gases: “Even so, the evaluation of the free energy (or entropy) of such a fluid at high densities by means of these nearly exact equations would take our best high-speed electronic computers (“magic brains”) many times longer than the currently accepted age of the earth.” In the 1970s, Seitz and Matsen (2) described “computer augmented lectures” in which textual results (graphics were not readily available then) from computations in atomic theory, spectroscopy, and kinetics were projected on a classroom screen. Computer hardware and software have evolved to where they are omnipresent in both national laboratories and undergraduate classrooms. Spreadsheets, Mathematica, MATLAB, python, and other software make mathematical problem solving and data visualization straightforward, thus allowing instructors and students to leverage the brain’s ability to process information through comparison and pattern recognition. Other chapters in this volume and articles in the chemical education literature detail the motivation for using computational resources to teach chemical concepts, and I won’t repeat the © 2019 American Chemical Society

Grushow and Reeves; Using Computational Methods To Teach Chemical Principles ACS Symposium Series; American Chemical Society: Washington, DC, 2019.

reasons here. Rather, I wish to focus on the design of in-class activities that make use of computing as a tool rather than a topic. The activities were developed with these considerations in mind: -

introduce computing as a flexible tool for insight into the stories behind data and for solving problems; use context and writing (narrative) to improve engagement with—and understanding of—chemistry concepts; apply assignment design strategies and frameworks that are known to positively impact student learning; help students develop high-level process skills such as information processing, problem solving, communication, and teamwork.

Assignments that use spreadsheets and computer algebra systems have been part of my physical chemistry courses for many years. These assignments were based on materials from the Symmath repository (3), articles in the Journal of Chemical Education, and textbook problems with modest changes to fit the needs or learning objectives for my students. When asked about which course assignments were “valuable for my learning,” students often ranked these assignments as “average” or “low” on course surveys. As a practicing chemist, I see the value of computing as a tool for understanding physical systems, and therefore sought to understand why novice chemists struggled to use the tool or failed to see its merit. Reviewing my course materials and student feedback carefully, I believe many of the problems can be attributed to three shortcomings: insufficient attention to context (making the content relevant to students), convolving content and process objectives, and too few opportunities for students to reflect upon, communicate, and assess their knowledge. In this chapter, I describe an assignment design framework that addresses these issues. Following an overview of computational notebooks and assignment design strategies, I share an approach for adapting elements of existing evidence-based frameworks to create what I call computational narrative activities (CNAs).

Computational Notebooks versus Computational Narratives An evolution of Knuth’s work on literate programming (4), Peréz defines literate computing as “the weaving of a narrative directly into a live computation, interleaving text with code and results to construct a complete piece that relies equally on the textual explanations and the computational components (5).” Initially developed in the late 1980s, the computational notebook instantiates literate computing. Computational notebooks are popular in proprietary applications such as Mathematica and Mathcad, and in open source applications like Jupyter (6). Because they can render text, mathematical equations, and multimedia sources, and evaluate code, computational notebooks have enabled authors to craft interactive textbooks (7), to provide supporting information for published articles (8), and to design interactive course materials (9, 10). In 2016, a Jupyter notebook demonstrating analysis of a black hole collision was published by the Nobel Prize-winning LIGOVIRGO team (11). (That fall, my physical chemistry students eagerly recreated an audio representation of a gravitational wave from a cosmic event that occurred 1.3 billion years ago.) Computational notebooks intended for use in physical chemistry courses are associated with textbooks (12, 13, 14), “computer friendly” chapter problems (15, 16), and found in online repositories (3). 164 Grushow and Reeves; Using Computational Methods To Teach Chemical Principles ACS Symposium Series; American Chemical Society: Washington, DC, 2019.

Existing computational notebooks can serve as excellent starting points for learning computational techniques and chemical concepts, but they vary widely in design choices and learning objectives. Many of these notebooks focus on mathematical problem solving, analysis, and visualizing results, with limited attention to communicating ideas or insights. In blog posts, Peréz and Granger (17) and Wolfram (18) argue the merits of computational narratives or computational essays as educational tools because of their utility in communicating scientific ideas. The pedagogical value of reflective writing for raising students’ curiosity, mastery of content, and knowledge retention is supported by research (19, 20, 21). Assuming the value of narrative in scientific analysis and communication, Rule and coworkers (22) sought to understand the use patterns of analysts, researchers, and educators who create computational notebooks. A corpus of nearly 1.3 million Jupyter notebooks written by 100,500 Github users was analyzed for the number and length of text and code cells (text cells contain free-form text; code cells contain programming statements that can be evaluated): 27.6% of the notebooks contained no text cells and 2.3% contained no code cells. Of notebooks containing text cells, word counts ranged from 0 to 55,000, with a median of 218. By far, the notebooks contained more lines of code than text. The researchers then filtered the corpus for notebooks written for academic purposes (e.g., explanatory analyses or tutorials) and found that these notebooks contain more lines of text than code, but only 3% had a text cell at the end of the notebook, suggesting there were few or no summary remarks. If the corpus is reflective of actual use, it seems that notebook authors are not exploiting the narrative aspect of computational narratives. Findings such as these, and my observations of students using computing in physical chemistry courses, have prompted me to move from computer-based assignments to computational narrative activities (CNAs).

Strategies for Designing Computational Activities My early attempts at using existing computational notebooks from external sources elicited student comments along the lines of, “Why are we learning programming instead of chemistry?” “Using the computer is harder than just doing the problem by hand.” “What does this have to do with the problems in the book?” To counter these sentiments, I explored ways to raise curiosity and sensemaking toward the activities. In How Learning Works, Ambrose, et al. (20), suggest ways to establish value in students’ perceptions of challenging coursework: -

connect the material to students’ interests provide authentic, real-world tasks show relevance to students’ current academic lives demonstrate the relevance of higher-level skills to students’ future professional lives identify and reward what you value show your enthusiasm for the subject

Comprehensive discussions of these strategies are available in instructor-friendly sources such as Bean (21), White (23), Kovacs and Sherwood (24), and the National Survey of Student Engagement Consortium (NSSE-WPA) reports (25). Common among these resources are recommendations to craft assignments that clearly address context, identify content and process objectives, and use writing to develop high-level thinking skills.

165 Grushow and Reeves; Using Computational Methods To Teach Chemical Principles ACS Symposium Series; American Chemical Society: Washington, DC, 2019.

The Value of Context in Student Learning The first four strategies in the list above can be addressed in activity design by considering a prominent element of student-centered pedagogies: context. Nentwig, et al. (26), provide direction for understanding context, “It is important in this understanding of chemistry teaching that the context is not a mere motivational trick in the beginning to lure the students into the chemistry. Nor is it an additive in the end to ‘further illustrate the subject matter’. The context is the red thread along which the investigation of the issue in question develops. It begins with the learners’ prior knowledge and experience, it is guided significantly by their questions and interest and it is linked to as many real world experiences as possible.” Relaying findings of a 15-year study of effective teaching practices, Bean quotes Ken Bain: “[confront students with] intriguing, beautiful, or important problems, authentic tasks that will challenge them to grapple with ideas, rethink their assumptions, and examine their mental models of reality” (21). Extensive background knowledge is not required for students to engage challenging problems if the problems are presented in a context that considers the student’s prior understanding: “The context is the starting point from where the teaching proceeds, and the ground principle is that the learner will start off from the context and then be aware of the content knowledge in demand to understand the issue in question on a ‘need-to-know’ principle [of Bulte, Westbroek, de Jong, and Pilot] (27).” An encouraging trend shows that textbook authors and publishers value the role of context in student motivation; several physical chemistry texts include vignettes explaining applications of concepts and problems that can convey to students the role of physical chemistry in biology, environmental science, and astronomy. Identify Content and Process Learning Objectives The fifth item in the list of strategies above involves identifying desired outcomes. The POGILPCL (28) guidelines recommend instructors distinguish content skills from process skills. Most instructors are comfortable identifying expected content objectives, but perhaps less so with process objectives. In a JCE editorial, Cooper implores educators to “say what we mean” (29) when telling students what they should learn, so definitions of process skills are necessary. The Enhancing Learning by Improving Process Skills in STEM (ELIPSS) project (30) provides definitions that can be helpful for developing activities (Table 1): Table 1. ELIPSS Project process skills definitions (30) Oral & Written Communication

Oral Communication: Exchanging information and understanding through speaking, listening, and non-verbal behaviors. Written Communication: Conveying information and understanding to an intended audience through written materials (paper, electronic, etc).

Teamwork

Interacting with others and building on each other’s individual strengths and skills, working toward a common goal.

Problem Solving

Identifying, planning, and executing a strategy that goes beyond routine action to find a solution to a situation or question.

Critical Thinking

Analyzing, evaluating, or synthesizing relevant information to form an argument or reach a conclusion supported with evidence.


Table 1. (Continued). ELIPSS Project process skills definitions (30) Management

Planning, organizing, directing, and coordinating one’s own and others’ efforts to accomplish a goal.

Information Processing

Evaluating, interpreting, and manipulating or transforming information.

Assessment (Self Assessment and Metacognition)

Self and Peer Assessment: Gathering information and reflecting on experiences to improve subsequent learning and performance. Metacognition: Thinking/reflecting about one’s thinking and how one learns, and being aware of one’s knowledge.

Along with these high-level skills, the activities should develop applied process skills—specific tasks students should be able to do. For example, students working a CNA that requires programming to model a phenomenon should be able to assign variables, define functions, and use flow control. A Framework for Designing Computational Narrative Activities The design of CNAs borrows elements from existing frameworks developed for effective writing and physical chemistry assignments. These frameworks are based on student-centered pedagogies and implement a learning cycle with phases related to exploring, inventing, applying and reflecting on new information. For example, POGIL materials are designed around a guided inquiry learning cycle with phases identified as Exploration → Concept Invention → Application: “[T]he exploration of a model occurs through direct questioning. Concept invention then takes place as students begin to see patterns and relationships in the data and terms are introduced. Finally, questions are posed that ask students to address the application of a concept to new situations (31).” In developing CNAs, I adapted some elements of the POGIL approach (32) in a way to leverage the capabilities of computational software and to emphasize the design strategies described in the Introduction. Computational narrative activities are intended to be worked in class by 2–3 person teams with facilitation by the instructor, and they contain following elements: -

Title: gives a sense of what the students will be learning, often phrased as a question Why? provides context, motivates learning content, describes real-world applications Prerequisites & pre-activity questions: conveys what prior knowledge or skills are assumed; sources for background information (e.g., textbook sections, video lecture, etc.) Learning objectives (LOs): communicate the process skills and content that the student should know after completing the activity Model: provides the “need to know” information associated with the LOs; can be in many forms including data, equations, graphical or aural representations, code snippets, or text; is the starting point of the learning cycle Inquiry questions: 3–5 questions designed to direct understanding, highlight key features of the Model, expose student misconceptions of the Model; questions target both process and content skills in LOs; prompt students for brief, narrative explanations Exercises and application tasks: straightforward applications of concepts being developed in the activity; may require judgment, decision-making, or delegation among team members Short reflections: prompt students to write narratives of the learning experience or thinking process, summarizes how the work came to be, asks for justifications for decisions, and


-

gives opportunity for self-assessment of confidence about LOs; short reflection prompts often follow inquiry questions and exercises. Problems and/or Long Reflections: engender situations where solutions are not immediately apparent, may have multiple parts; may extend original context or be framed in a different context to assess knowledge transfer; may require synthesis of concepts (perhaps from other assignments); judgment or assumptions with justifications); often assigned as homework (out-of-class).

Typically, each “chapter” in a notebook activity consists of a model, 3-5 inquiry questions, 1-3 exercises, 1-3 short reflections and 1-2 long reflections or problems. Early in the term when students are developing their python skills, the in-class portion of a single chapter can be completed in a 1hour class period if long reflections and problems are assigned as homework. After students have facility with python, it is feasible to introduce more than one chapter. CNAs with several chapters require 2-3 class periods or a 3-hour laboratory period. In the next section, excerpts from existing CNAs illustrate how the various elements listed above can address the design goals.

Examples of Computational Narrative Activities Several CNAs were developed to introduce concepts from physical chemistry and from the python programming language. My P-Chem courses are structured around the POGIL pedagogy, and students are familiar with the practices of forming teams, assigning roles, and managing in-class activities. An overview of the context (personal/real-world connection), computing (applied process) skills, and chemistry content covered in several CNAs is given in Table 2. Table 2. Context, computing skills, and content areas addressed in several computational narrative activities Title/Context

Computing Skills

Content

Why do stars have different colors? Importing and massaging data, Black body radiation, Planck Relationships between non-linear curve fitting, numerical relationship, stellar emission integration, plotting spectroscopy, energy density temperature, color, and energy Cricket thermometers: How can we estimate the energy of a metabolic process?

Data linearization, linear regression, non-linear curve fitting, error analysis

Integrated rate laws, Arrhenius equation, activation energy, metabolic process

How does a butane lighter work? Why gases condense to liquids

Plotting, visualization, extrapolation, error analysis

Equations of state, intermolecular forces, compressibility

Are double bonds double strength? Models for bonds

Correlating model parameters, using ab intio computational chemistry package

Morse potentials, vibrational and dissociation energies, bonding models

Can foam height distinguish Importing data, differential beers? Developing a kinetic model equations, numerical integration

Integrated rate laws, error in fitting parameters

Below are excerpts showing various elements from the CNAs. The figures are screen captures of Jupyter notebooks running a python 3.6 kernel. Students have little-to-no prior background in python programming; python concepts are typically introduced in the activity as a worked example or a directed task (see the section Using Computational Narrative Activities in Physical Chemistry 168 Grushow and Reeves; Using Computational Methods To Teach Chemical Principles ACS Symposium Series; American Chemical Society: Washington, DC, 2019.

Courses). Jupyter code cells contain python statements that are evaluated by the python interpreter, and markdown cells (the equivalent of a text cell) render content as rich-text (a subset of HTML that provides formatting, hyperlinks, embedded images, etc.). Mathematical typesetting is accomplished using standard LaTeX markup. Title and Introduction Provide Context and Pique Interest Presenting the title or a key concept in the form of a question can raise curiosity and provide context. Images or questions that connect content to personal experience typically elicit casual storytelling among students, which suggests they are engaged with the topic. Figure 1 shows the introductory material for Cricket Thermometers, a CNA exploring activation energy in the context of insect stridulation (chirping). The Jupyter audio player allows students to hear and measure cricket chirp rates at different temperatures. Later in the CNA, they plot a spectrogram and compare the precision of measurements with the ear and a stopwatch to a digital recording with a sampling resolution of 44KHz. A homework problem has them take fast Fourier transforms of the recordings with python to determine chirp frequencies.

Figure 1. Markdown cells and code cells with title, introduction, an image, and audio player. The following screenshots come from the CNA, “Why do stars have different colors?” which introduces students to the Rayleigh-Jeans and Planck models for black body radiation. In this activity, students use python to fit stellar emission spectra (33) to the Planck model and estimate the temperatures of several stars. Figure 2 shows the title and introduction which explain the context and learning objectives of the activity.


Figure 2. Title, introduction, and learning objectives for a CNA exploring black body radiation in context of stellar emission spectra. Prerequisites and Pre-Activity Questions Communicate Expected Background Knowledge In this section of a CNA, background skills and concepts are listed so students know what background knowledge is assumed. Pre-activity questions may be assigned before or at the beginning of a class session, and prompt students to reflect on their prior knowledge and/or perceptions of the CNA’s context. Low cognitive load writing prompts such as listing facts, writing personal questions about a topic, or describing personal experiences help student make connections and activate prior knowledge. For example, in the CNA “Why do stars have different colors?” pre-questions include, “List several things you know about stars and several questions you have about stars.”, “How hot is a star? What is your reasoning (even if it is ‘wild guess’)?”, “Speculate on the origins of the colloquialisms ‘red hot’ and ‘white hot’.” The cells shown in Figure 3 convey the python-related background skills and libraries that will be used in the CNA. A Jupyter markdown cell renders the formatted text and the code cell (shaded region) evaluates python statements.

Figure 3. Python background skills (with links to sources) and libraries that will be used in the CNA. 170 Grushow and Reeves; Using Computational Methods To Teach Chemical Principles ACS Symposium Series; American Chemical Society: Washington, DC, 2019.

Early in the course, students are introduced to python import statements. In python-based activities, import statements implicitly signal prerequisite python knowledge. If the numpy and matplotlib libraries are imported, students surmise they should be able to manipulate numpy arrays and create plots, which are skills that have been developed in other assignments. Annotating the imports with comment statements (lines beginning with #) can clarify the expectations. This section may also contain links to information on prerequisite python topics that have not been covered, or topics for which additional instruction may be useful. The excerpt shown in Figure 3 has links to a video tutorial demonstrating the pandas library for importing and plotting data from a spreadsheet file. Model or Information The Model or Information section is a central part of the activity. Similar to the approach taken in other POGIL materials (34, 35), a Model is designed to be explored in self-managed teams, can take just about any form that conveys information (a plot, a table, an image, an equation), is engaging, and uses standard vocabulary with concise, focused language. The Model in Figure 4 introduces the Morgan-Keenan stellar classification system which students later use to estimate the temperature of several stars in the Milky Way galaxy. Also included are links to supplementary biographical and historical information (additional context) related to the Model.

Figure 4. A Model introducing a stellar classification system, with additional information available through hyperlinks. CTQ 3 prompts students to apply the Model to a recognizable example. Figure 5 shows how the Planck model for black body radiation is introduced. Prior to the Planck model, students have seen and responded to questions related to the Rayleigh-Jeans model the ultraviolet catastrophe. A Model can also take the form of a worked example or problem as shown in Figure 6. The Model in Figure 6 requires no interaction on the part of the student until the inquiry or exercise/application section which is encountered later in the activity.


Figure 5. Introducing the Planck model for black body radiation.

Figure 6. A model presented in the form of a worked example. 172 Grushow and Reeves; Using Computational Methods To Teach Chemical Principles ACS Symposium Series; American Chemical Society: Washington, DC, 2019.

Inquiry Questions, Exercises, and Application Tasks Inquiry questions, exercises, and application tasks steer students through the phases of the learning cycle. For example, after introducing the Planck and Rayleigh-Jeans curves shown in Figure 5, several inquiry questions ask students to identify similarities and differences in the curves, and to articulate patterns in Planck model that comport with the requirement that total energy of a blackbody be finite. Following the exploration of concepts, exercises or application tasks have students make predictions about the behavior of blackbody radiators. Students then write python code to verify their predictions and extend the scope of their exploration over a temperature range spanning cosmic background radiation (a few Kelvin) to Type O star emission (3×105 K). An exercise from this activity is shown in Figure 7.

Figure 7. An exercise with a prompt for narrative. The optional task provides additional challenge for students wishing to develop python skills. Questions that ask teams to predict and verify outcomes of a simulation or model function are useful for developing information processing, communication, teamwork, and critical thinking skills; because the CNA is worked in class, evidence of desired behaviors associated with these skills can be monitored. The narrative aspect of CNAs is again emphasized by reminding students their responses should consist of complete sentences, be of an appropriate length, and provide defensible explanations, justifications, or interpretations. A goal of the narrative is to capture the thinking that led to the conclusion. Frequently, unexpected results or divergent explanations arise; reconciling initial expectations with contradictory evidence is true learning. Writing about moments of “cognitive dissonance” will help students identify areas that deserve further study. Some exercises or questions have the team delegate tasks. Each team member or subgroup completes the computing task for, say, different starting conditions. The results are pooled, and additional questions encourage students, for example, to look for patterns, anomalies, or limits of a model equation. Potential pitfalls during this phase of working the CNA include perfunctory predictions or answers. Furthermore, access to internet resources may lead to students looking up answers rather 173 Grushow and Reeves; Using Computational Methods To Teach Chemical Principles ACS Symposium Series; American Chemical Society: Washington, DC, 2019.

than working through the tasks. I have found that temporarily allowing teams to divide into twoperson groups that reassemble into the full team helps keep all members engaged. Indeed, when the teams come back together, students get an extra chance to explain their findings before reporting to the class. Reviewing my early attempts at writing CNAs, and reviewing other instructional notebooks, I noticed too many convoluted questions or exercises that require students apply computing, chemistry, and analysis skills in the same task. Again, considering the specific behaviors or results each task should elicit can lower confusion and improve transparency of the assessment process. Prompts for Written Responses The case for narrative as a means to improve students’ mastery and retention of knowledge drives the frequency of prompts for written responses in CNAs. To provide variety in cognitive load and maintain interest, writing prompts are adapted from the conceptual hierarchy described by Kovac and Sherwood (24) and summarized in Table 3. Table 3. Conceptual hierarchy of writing tasks Level

Form

Low

listing, definition, seriation

Mid

classification, summary, compare/contrast

High

analysis, present academic/scientific argument

Low-level forms appear in a CNA as part of the in-class work and in homework assignments (e.g., Exercise 4(b) in Figure 7). Mid-level forms are excellent for providing closure to a CNA, giving both the student and the instructor a sense of the student’s progress toward the expected outcomes. High-level forms are rarely part of an in-class CNA, but they do serve as the basis for essay and project assignments and may bridge class work and laboratory experiments. The closing reflection of a CNA should also encourage self-assessment. I remind students reread the objectives at the beginning of the activity and then assess their confidence in meeting the objectives. I also ask for reasons or evidence supporting their assessment. Finally, a one or two paragraph personal narrative can communicate the student’s perceptions of the main ideas, the most challenging tasks/concepts, and the most interesting or rewarding aspects of working the activity. These narratives are not graded beyond “completion with evidence of thoughtful reflection.” I have found personal stories to be quite valuable for identifying common interests or concerns shared by the group which I can then acknowledge in subsequent class meetings. Approaches for Assessing and Evaluating Student Work With the emphasis on the narrative aspect of CNAs, I appreciate assessment methods that are time-efficient and useful to students. A holistic scoring guide (23, 24) meets these needs. The scoring guide describes thresholds for receiving scores (1–6) in areas of content, development, organization/ style, and mechanics. I recently became aware of the ELIPSS rubrics and intend to use these for assessing the skills listed in Table 1. The collection of VALUE (36) rubrics is another source of vetted instruments for instructors seeking to assess outcomes associated writing, problem-solving, and teamwork. 174 Grushow and Reeves; Using Computational Methods To Teach Chemical Principles ACS Symposium Series; American Chemical Society: Washington, DC, 2019.

An approach to developing self-assessment skills is to provide a Jupyter notebook that provides expected answers for exercises, working code examples, and a list of key points that should be included in free-response questions. After distributing the solution notebook, I ask students to compose a concise summary that compares their work to the solution notebook and describes their evidence of meeting the learning objectives. I specifically ask them to explain how mistakes diverged from the given solution and how the errors can be corrected. Even developing students are able to identify incorrect answers or weak responses after comparing their work to an exemplar (24). This commentary is evaluated, not the answers, so students are rewarded for critically analyzing their own work.

Using Computational Narrative Activities in Physical Chemistry Courses CNAs have been used in the physical chemistry courses and laboratory at Coe College for two years. Coe is an undergraduate liberal arts institution with an ACS-certified chemistry department. Physical Chemistry I is required for the basic chemistry major. Physical Chemistry II is recommended for all majors, but required of only those seeking ACS certification. The majority of students enrolling in P-Chem are chemistry majors with the remainder being physics majors and students interested in materials science or engineering graduate programs. Prerequisites for Physical Chemistry include two semesters of calculus and two semesters of general physics; recommended courses include Modern Physics, Linear Algebra, Calculus III, and Differential Equations. It is common for students to enter the college with Advanced Placement credit and they usually opt to take the calculus and physics prerequisites early in their undergraduate careers. Often students have not engaged concepts from these courses for more than a year, and a common struggle is relearning concepts and skills that have been dormant for some time (10). There is no major requirement for a programming course, and typically one or two students will have had some exposure to the python programming language prior to taking P-Chem. According to instructors in the Mathematics and Physics departments, students may have cursory exposure to math software (Sagemath or Mathematica) in Calculus II or General Physics, but it is not used as a teaching/learning tool. P-Chem courses meet for 3 one-hour class sessions, and 1 three-hour laboratory session per week over a 15-week semester. The primary pedagogy is POGIL, with about 80% of class and laboratory time dedicated to student-centered activities in small groups. The same practices are used for in-class CNAs. Smaller class sizes allow me to interact with all teams each period. Instructors with larger classes would likely benefit from teaching assistants or veterans of the course trained in facilitating in-class activities. Content is distributed between Physical Chemistry I and II using the 2015 ACS-CPT model of foundational and in-depth coursework. Rather than a traditional semester-long treatment of quantum chemistry/spectroscopy and thermodynamics/kinetics, all topics are covered in both semesters. Physical Chemistry I consists of several modules covering kinetics, one-dimensional quantum systems with introductory spectroscopy, and classical thermodynamics. Topics such as data and error analysis, using computational chemistry software, and python programming are introduced ad hoc during classroom or laboratory activities. Physical Chemistry II modules revisit several of concepts introduced in the first semester, developing depth and rigor in quantum mechanics (two- and three-dimensional quantum models), computational chemistry theory, spectroscopic analysis, and statistical thermodynamics. Through multiple exposure to general topics of quantum, spectroscopy, thermodynamics, and kinetics in several different contexts, students 175 Grushow and Reeves; Using Computational Methods To Teach Chemical Principles ACS Symposium Series; American Chemical Society: Washington, DC, 2019.

hopefully recognize patterns in computational methods and minimize compartmentalization of knowledge. A common physical chemistry textbook is used in the courses. The textbook serves primarily as a reference and a source of homework problems. As mentioned, CNAs often include cross-references or links to recommended reading, detailed derivations, worked examples, or additional practice problems. Introducing Python and Accessing Jupyter in the Classroom Surprisingly little formal instruction in python is necessary for students to grasp and use the language. I attribute this to the straightforward design and syntax of the language, and excellent tutorials available online. Early in the term, I dedicate one laboratory period to familiarizing students with the Jupyter ecosystem, python language basics, and submitting assignments. Python data structures, flow control, and libraries are introduced as needed for specific CNA tasks. Usually a worked example or short sequence of inquiry questions is sufficient for a few students to grasp the programming concept; they then tutor their teammates on its use. As mentioned above, CNAs include a Prerequisites section that identifies expected python skills and provides links to appropriate tutorials. Jupyter installation details are readily available through the project’s website, www.jupyter.org. Students wishing to run Jupyter on their own laptops have little difficulty installing the open source Anaconda scientific python distribution from www.anaconda.com. Jupyter notebooks are simple json files, so any mechanism capable of distributing and receiving text files (course management system, shared cloud storage, Github, etc.) can be used in the course. The web service CoCalc, www.cocalc.com, provides access to a complete linux-based computing environment through a browser. Students with CoCalc accounts have access to Jupyter and many other computing tools such as Sagemath, LaTeX, and more than a dozen programming languages. The lead developer of CoCalc is a mathematics professor and has provided features designed for classroom administration like assignment distribution and collection, grading and feedback tools, and real-time chat. CoCalc mimics two compelling features available in Google Docs: collaborative editing of Jupyter notebooks and TimeTravel which allows authors to “rewind and replay” editing changes. Although free accounts are available, they are given low priority in the service queue and students occasionally complain about slow response during peak use times. I recommend purchasing basic service accounts for the entire class which currently cost less than 20 USD per student per semester. Paid accounts receive top priority which is useful when thousands of people are simultaneously using CoCalc. Jupyter Hub (the Project Jupyter multi-user server) and CoCalc are both open source projects and can be hosted on local servers if desired. Regardless of the whether students use their own laptops or a distributed service, I have experienced only minor technical problems using Jupyter notebooks in the classroom.

Observations After Using CNAs in the Classroom The evolution of CNAs has been organic and thus far has lacked a formal plan for measuring learning gains. I will, however, share my observations on some of the successes and challenges of using CNAs to teach physical chemistry concepts.


-

-

Evaluating content skills: Because of the mix of topics covered each semester, I select appropriate questions from ACS Physical Chemistry exams to create a final exam for the course. To my knowledge, the ACS does not provide an item analysis for exam questions, so comparisons to national test results is dubious. However, for a general sense of student performance, the class exam averages correlate to the 70th percentile nationally, consistent with scores earned before the introduction of CNAs. Hence, it is not possible to claim CNAs have improved exam scores, but there is no evidence that their use (with a commensurate loss of time on other activities) has lowered exam scores. In other evaluations such as quizzes with questions from traditional sources, students are performing as well as with print-form curricular materials. Other class assignments and laboratory analyses require that python be used, and students have been able to able to use this tool effectively to solve problems that would not be solvable using different methods. Assessing process skills: Coe administers course surveys that include items very similar the process skills listed in Table 1. In the survey, students rate (1–5) how the course improved their learning or abilities in Problem Solving, Using Information, Communication Written and Oral, Using Technology, and Working in Teams. In Physical Chemistry courses, class average ratings for all of these areas consistently exceed the college mean except for Communication Oral which matches the mean. Since implementing CNAs, the amount and quality of student-generated prose has increased measurably. Word counts typically exceed 500 words for a completed assignment (about 2 typewritten pages). Traditional paper-based assignments that do not emphasize narrative require considerably less student writing. Accepting the adage “good writing is good thinking” (which is why both are so hard), students can show evidence of the headway they are making in thinking skills.

With the ability to intertwine computer code and text, to provide instant feedback, and to create tables or graphics, computer notebooks have been helpful for improving students’ ability to “think mathematically” (37). Based on student feedback, a welcome side effect of CNAs is lower math anxiety (many entering physical chemistry students express low confidence in their calculus and algebra skills). Although the habits of mind needed for computing and mathematics are similar, apprehensive students seem receptive to learning basic programming instead of relearning calculus. Conversely, students with confidence in their math skills instantly see the value in solving problems with the computer, and many report using python or Sagemath in other courses. -

-

Student perceptions of CNAs: I ask students to list things they like or value about CNAs and things they dislike or would change about CNAs. Representative comments for things most commonly liked/valued: “Learning the chemistry behind [stars, beer, crickets, etc.]” (the context), “It forces me to learn something well enough to write about it,” “Being able to put videos and equations in the notebooks,” “Knowing that python can do stuff, even if I don’t know how to make python do it.” Common dislikes include, “It’s frustrating for the group when it takes us forever to get python to work right,” “Too much writing for a chemistry class,” “I’d rather learn about chemistry than learn about computers.”On the positive side, several alums report that the introduction to computing and professional communication were valuable to their education and career (perhaps more so than the physical chemistry knowledge?). I am still experimenting with workflows for assessing and giving feedback on notebooks. It can be difficult for students to find cells containing feedback if the instructor’s cells are intermingled with other cells. Currently, I type feedback into markdown cells with a large, 177 Grushow and Reeves; Using Computational Methods To Teach Chemical Principles ACS Symposium Series; American Chemical Society: Washington, DC, 2019.

-

-

-

-

-

colored typeface and return the “graded” notebook to the student. To avoid a manual approach such as this, I recommend instructors investigate nbgrader which is a Jupyter sub-project that provides automated grading for code cells. It uses the python testing framework to compare a student’s code output to an expected result. An automated grading system should work well for CNA tasks involving code and numerical answers, but free-response questions and narratives must be read and scored manually. Another Jupyter project, nbdime, shows two notebooks side-by-side with differences highlighted, and would appear to be a good solution. However, using nbdime to show a submitted notebook next to a graded notebook has proved confusing to students because nbdime is currently a command-line application. Direct instruction on using nbdime may make it a viable tool for providing feedback in notebooks. A compelling feature of Jupyter is the collection of widgets and extensions (38) that support many digital forms including audio, video, interactive maps, and 3-D volume rendering (particularly useful for visualizing multivariate functions). With these tools, Models and Exercises can contain rich artifacts that go beyond static equations or plots. Data can be imported directly from curated repositories or students’ own measurements, strengthening connections between the activity and real world problems. Furthermore, Models containing code cells can hint at professional communication skills. For example, rather than presenting a textbook plot, which hides lots of data massage, normalization, or reduced units, it is possible to show the actual code used to create a professional looking plot. In contrast to paper-based POGIL activities, when working on CNAs, 4-person teams often break into 2+2 person or 1+3 person subgroups for periods during the class. The smaller groups continue to interact, but behaviors around a computer are different than those with printed activities. As always, it is important for team managers and the instructor to watch for students succumbing to internet distractions or disengaging for extended periods. CNA questions that prompt students to “Discuss with your team…” remind students to re-engage with the group. Students enthusiastically use the linking, formatting, and embedding capabilities of markdown cells in their responses. This capability can be a benefit in finding connections among sources, but also a liability when the technology distracts from reflection (it’s not uncommon for students to include visually stimulating artifacts that do not communicate understanding of chemistry). Over half of those enrolled have chosen to use Jupyter for lab write-ups, and a small number report using it in other courses. Several students mentioned value in learning how to use LaTeX markup for typesetting mathematical forms. As with hand-written assignments, I seem to regularly admonish students for deleting work that contain mistakes. When asked why they want to expunge mistakes, a common response is, “I don’t want to remember bad information.” Unfortunately, this signals a student is planning to rely on memorization rather than understanding the origin of the mistake. To counter this, I reward students for keeping evidence of “good mistakes” if they include a narrative passage identifying the mistake and the error in thinking that led to the wrong result. Python has libraries for many communication interfaces (USB, RS-232, 488, I2C, etc.). Students use python and Jupyter notebooks to control laboratory instruments and plot or manipulate data in real time. A specific laboratory example: after completing the CNA on black body radiation, students record the solar spectrum using an interfaced Ocean Optics 178 Grushow and Reeves; Using Computational Methods To Teach Chemical Principles ACS Symposium Series; American Chemical Society: Washington, DC, 2019.

spectrometer and fit it using the model function developed in the “Why do stars…” CNA. Students confirm their prediction of the Sun’s temperature with data collected themselves. Students come to see Jupyter/python as a broadly useful tool for both class work and lab work.

Conclusions The computational narrative activity is a design concept for writing interactive assignments conducive to a student-centered pedagogy. A CNA is developed using elements from researched design frameworks, and emphasizes context, narrative (writing to learn), content skills development, and process skills development. Written using the Jupyter ecosystem, CNAs introduce python programming as a tool for gaining insight into scientific problems and communicating ideas. Upon completing the activities, students have evidence they are meeting learning objectives such as the ability to communicate scientific ideas, to model complex systems, to use computing methods to solve numerical and symbolic problems, and to create professional visual representations and documents. Although the examples presented in this chapter use Jupyter and python, the design considerations for CNAs can be adapted for use with other computing environments. Combining the power of these systems with good assignment design helps students utilize computing as a tool to gain and communicate insight.

References 1.

Scott, R. L. Some Unsolved Problems of Liquids and Solutions. J. Chem. Educ. 1953, 30, 542–549. 2. Seitz, W. A.; Matsen, F. A. Computer Augmented Lectures. J. Chem. Educ. 1974, 51, 192–193. 3. Zielinski, T. J. Fostering Creativity and Learning using Instructional Symbolic Mathematics Documents. J. Chem. Educ. 2009, 86, 1466. 4. Knuth, D. E. Literate Programming; Center for the Study of Language and Information: Stanford, CA, 1992. 5. Perez, F. “Literate computing” and computational reproducibility: IPython in the age of data-driven journalism. http://blog.fperez.org/2013/04/literate-computing-and-computational.html (accessed July 1, 2018). 6. Project Jupyter.https://jupyter.org/ (accessed July 01, 2018). 7. Vanderplas, J. T. Python Data Science Handbook; O’Reilly Media: Sebastopol, CA, 2016. 8. Shen, H. Interactive Notebooks: Sharing the Code. Nature 2014, 515, 151–152. 9. Barba, L. AeroPython. https://github.com/barbagroup/AeroPython (accessed July 1, 2018). 10. Goss, L. M. In The Flipped Classroom Volume 1: Background and Challenges; Muzyka, J. L., Luker, C. S., Ed.; ACS Symposium Series 1223; American Chemical Society: Washington, DC, 2016; pp 29−54. 11. Vallisneri, M.; Kanner, J.; Williams, R.; Weinstein, A.; Stephens, B. The LIGO Open Science Center. J. Phys. Conf. Ser. 2015, 12021. 12. Cropper, W. H. Mathematica Computer Programs for Physical Chemistry; Springer-Verlag: New York, 1998.


13. Noggle, J. Physical Chemistry Using Mathematica; HarperCollins College Publishers: New York, 1996. 14. Metiu, H. Physical Chemistry Quantum Mechanics; Taylor and Francis: New York, 2006. 15. Engel, T.; Reid, P. Physical Chemistry, 3rd ed.; Pearson: Boston, MA, 2013. 16. Atkins, P. W.; de Paula, J. Physical Chemistry, 10th ed.; Oxford University Press: New York, 2014. 17. Pérez, F.; Granger, B. Project Jupyter: Computational Narratives as the Engine of Collaborative Data Science. https://blog.jupyter.org/project-jupyter-computational-narratives-as-theengine-of-collaborative-data-science-2b5fb94c3c58 (accessed July 1, 2018). 18. Wolfram, S. What Is a Computational Essay? http://blog.stephenwolfram.com/2017/11/ what-is-a-computational-essay/ (accessed July 1, 2018). 19. Anderson, P.; Anson, C.; Gonyea, B.; Paine, C. Summary: The Consortium for the Study of Writing in College (CSWC). http://compile.org/wpa+nsse/ (accessed July 1, 2018). 20. Ambrose, S. A.; Bridges, M. W.; DiPietro, M.; Lovett, M. C.; Norman, M. K.; Mayer, R. E. How Learning Works: Seven Research-Based Principles for Smart Teaching; Josie-Bass: San Francisco, CA, 2010. 21. Bean, J. C. Engaging Ideas: The Professor’s Guide to Integrating Writing, Critical Thinking, and Active Learning in the Classroom; Josie-Bass: San Francisco, CA, 2011. 22. Rule, A.; Tabard, A.; Hollan, J. Exploration and Explanation in Computational Notebooks. ACM CHI Conference on Human Factors in Computing Systems. http://hal.archives-ouvertes.fr/hal01676633 (accessed July 1, 2018). 23. White, E. M. Assigning, Responding, Evaluating: A Writing Teacher’s Guide; Bedford/St. Martin’s: Boston, MA, 2007. 24. Kovac, J.; Sherwood, D. W. Writing Across the Chemistry Curriculum: An Instructor’s Handbook; Prentice-Hall: Upper Saddle River, NJ, 2001. 25. Anderson, P.; Anson, C.; Gonyea, B.; Paine, C. The Contributions of Writing to Learning and Development: Results from a Large-Scale Multi-institutional Study. Res.Teach. Eng. 2015, 50, 199–235. 26. Nentwig, P. M.; Demuth, R.; Parchmann, I.; Ralle, B.; Gräsel, C. Chemie im Kontext: Situating Learning in Relevant Contexts while Systematically Developing Basic Chemical Concepts. J. Chem. Educ. 2007, 84 (9), 1439–1444. 27. Broman, K.; Parchmann, I. Students’ Application of Chemical Concepts When Solving Chemistry Problems in Different Contexts. Chem. Educ. Res. Pract. 2014, 15, 516–529. 28. Hunnicutt, S. S.; Grushow, A.; Whitnell, R. Guided-Inquiry Experiments for Physical Chemistry: The POGIL-PCL Model. J. Chem. Educ. 2015, 92, 262–268. 29. Cooper, M. M. It Is Time To Say What We Mean. J. Chem. Educ. 2016, 93, 799–800. 30. Cole, R.; Lantz, J.; Ruder, S.; Reynders, G.; Stanford, C. Enhancing Learning by Improving Process Skills in STEM. http://www.elipss.com (accessed July 1, 2018). 31. POGIL Project. https://pogil.org/educators/additional-resources (accessed July 1, 2018). 32. Moog, R. S.; Spencer, J. N. Process Oriented Guided Inquiry Learning (POGIL); American Chemical Society: Washington, DC, 2008; pp 1−13. 33. Pickles, A. J. A Stellar Spectral Flux Library: 1150-25000A. Pub. Astron. Soc. Pac. 1998, 110, 863–878. 180 Grushow and Reeves; Using Computational Methods To Teach Chemical Principles ACS Symposium Series; American Chemical Society: Washington, DC, 2019.

34. Spencer, J. N.; Moog, R. S.; Farrell, J. J. Physical Chemistry: A Guided Inquiry Thermodynamics; Houghton-Mifflin: Boston, MA, 2004. 35. Shepherd, T. D.; Grushow, A. Quantum Chemistry and Spectroscopy: A Guided Inquiry; Wiley: Somerset, NJ, 2014. 36. Rhodes, T. Assessing Outcomes and Improving Achievement: Tips and Tools for Using the Rubrics; Association of American Colleges and Universities: Washington, D.C., 2009. 37. Schoenfeld, A. H. Learning to Think Mathematically: Problem Solving, Metacognition, and Sense Making in Mathematics. J. Educ. 2016, 196, 1–38. 38. Jupyter Interactive Widgets. https://jupyter.org/widgets (accessed July 1, 2018).


Computational Narrative Activities: Combining Computing, Context

Recommend Documents