Teaching basic lab skills
for research computing

Our IUSE Proposal Was Rejected

We got word a few days ago that our proposal to the NSF's Improving Undergraduate STEM Education program had been rejected. The panel summary agreed that software training is a good idea, but were not convinced about our plans to shift from training grad students to undergrads. In particular, they were not convinced that Unix-based workshops would be best for undergrads, and felt that not being embedded in the regular curriculum was a weakness. These are fair criticisms:

  1. Most undergraduates only use GUIs and cannot navigate the terminal to save their lives. The perception that Software Carpentry is Unix-specific therefore hurt the proposal, and showed that we didn't clearly explain our focus on underlying concepts.
  2. The fact that we wouldn't be part of the regular curriculum is more difficult to address. The fact is, it really is hard to get undergrads to do things that interfere with earning grades, just as it's often hard to get grad students to do anything that doesn't immediately lead to a publication. Some of their concern about impact also seemed to be due to us concentrating on REU students, a self-selected bunch that already are motivated.

At the same time, though, the panel did not connect the results from years of Software Carpentry workshops and the expected impact of this effort. This indicates that the proposal did not effectively communicate how well our experience to date has laid the groundwork for efforts like this.

We're obviously disappointed by this rejection, but we've some useful lessons, and we hope that they will inspire others to put forward proposals of their own.


The Impact of Intensive Software Skills Training on Students' Scientific Careers

Rachel Slaybaugh, Kaitlin Thaney, Lorena Barba, C. Titus Brown, and Paul Wilson
with Ethan White, Tracy Teal, Greg Wilson, and Kathryn Huff

The Problem: Computational Competence in Science

Scientists and engineers invented electronic computers to accelerate their work, but two generations later, many researchers in science, technology, engineering, and mathematics (STEM) are still not computationally competent: they do repetitive tasks manually instead of automating them, develop software using a methodology best summarized as "copy, paste, tweak, and pray" and fail to track their work in any systematic, reproducible way.

While the World-Wide Web was created by a scientist to help his peers share information, many still use it primarily as a way to find and download PDFs. Researchers may understand that open data can fuel new insights but often lack the skills needed to create and provide a reusable data set. Equally, any discussion of changing scientific publishing, making research reproducible, or using the web to support "science as a service" must eventually address the lack of pre-requisite skills in the general STEM research community.

Studies have repeatedly shown that most researchers learn what they know about computing by word of mouth, but this approach is failing to meet present-day needs: most faculty would agree that today's graduates are no more able to use computing and the web in their research than they were a generation ago. Attempts to integrate more training in basic computing skills into undergraduate education have largely failed to take root for several reasons:

  1. The curriculum is full. Undergraduate STEM programs already struggle to cover material regarded as core to their field. While many scientists would agree that more material on programming, reproducible research, or web-enabled science would be useful, there is no consensus on what to take out to make room.
  2. The blind leading the blind. Many faculty lack computational skills themselves and hence are unable to pass them on.
  3. Cultural difference. Scientists and software developers have different priorities and different approaches to problem solving, which often impedes collaboration and knowledge transfer.

One final issue is that the rewards are unknown. Open, web-based science is still in its infancy, so there is no general understanding of what people might need to know in order to incorporate it into their research careers. Since it is hard to measure something if you don't know what to look for, or if it is so young that there hasn't actually been long-term impact, little systematic study has been done to date of whether early training in the skills needed for this new kind of science actually has an impact, and if so, how and how much. Without such feedback, there is no systematic way to improve the training programs that currently exist.

Proposed Work: Leveraging Proven Curriculum to Promote Computational Competence in Science

This proposal builds on the success to date of the Software Carpentry workshops, a proven curriculum of essential software skills that enhance the productivity of graduate students, post-docs, and faculty. We propose to:

  1. conduct formative evaluation of the impact of software skills training for undergraduates likely to continue in research careers as they progress through the early stages of those careers;
  2. conduct summative evaluation of the training's overall impact on a multi-year timescale in order to improve the content and presentation of the training; and
  3. disseminate the resulting curriculum widely.

We will run software-skills workshops for undergraduate students taking part each year in summer research opportunities such as the NSF's Research Experience for Undergraduates (REU) program, at or near the start of those students' time in the lab. Based on data already collected from Software Carpentry workshops, we expect this training will help them be more productive during their research (graduate-level participants in our existing workshops typically report that what we teach saves them a day per week) and will prepare them to work in a world where all aspects of science are increasingly dependent on computing.

These undergraduates will serve as the treatment population for a five-year study of the impact of this training on their careers in general, and their involvement with open and web-enabled science in particular. In order to conduct this study, we will hire an expert in educational assessment, whose full-time work for the duration of the project will be to explore the effects of the training on workshop participants.

Workshops: A Distributed Model for National Impact

We will run two-day workshops at a steadily increasing number of sites each year for five years, timed to coincide with the start of the summer influx of undergraduate research students. Each workshop will be offered to a minimum of 40 learners per site, giving us a target study population of at least 2200 students by year 5. The content will be tailored to meet local needs but will be based on what is being used at that time by Software Carpentry and affiliated educational efforts. By design, it is straightforward to adapt workshop materials and contribute changes to the Software Carpentry course material. These features enhance the portability and flexibility of the workshops and increase the likelihood of wide dissemination beyond this project.

The home sites for investigators named in this proposal (George Washington University, Michigan State University, University of California, Berkeley, University of Wisconsin – Madison, and Utah State University) will run workshops in each of those years. Other sites will be added each year, expanding the total to 15 in year 5 and accordingly increasing the size of our study population. We will focus expansion on NSF REU sites but, as detailed below, we will also offer some workshops to other communities.

One set of possible sites for expansion are those campuses included in the "Condo of Condos" consortium, recently recommended for funding by the National Science Foundation. The Software Carpentry workshops proposed here will be valuable to that consortium in meeting its goals of increasing the number and diversity of researchers using advanced cyberinfrastructure and of developing data science practitioners.

Beyond this consortium, we will recruit sites for hosting workshops by identifying locations at which we could have the largest impact and/or that contribute most to our goal of increasing diversity. If we find there are more sites interested in hosting workshops than we are able to support in a given year, we will select the subset of sites that best meet our goals.

Curriculum: From Tools to Techniques to Concepts

While there is considerable scope for customizing workshops to accommodate learners' prior experience and discipline-specific needs, what these workshops seek to convey is the best practices a researcher needs to be computationally competent:

  • how to create, use, and share structured data
  • how to automate repetitive tasks;
  • how to track and share work over the web; and
  • how to grow a program in a modular, testable, reusable way.

With these objectives, the base workshop format will be divided in to four modules. All workshops will be hands-on so that learners can "learn by doing" and have experience with concrete examples. A typical workshop will devote roughly half of a day to each of the following:

  1. Working with Data: this module will introduce students to efficient data manipulation. Learners will work with data representative of their field of study. By the end of this module, learners will have the basic skills for parsing data, using and structuring databases, and conducting more sophisticated statistical analyses. We will point out why spreadsheets are insufficient for these types of tasks.
  2. Automation with the Unix Shell: this module will introduce learners to the shell; teach how to view, search, and manipulate text files at the command line; and introduce basic automation at the command line.
  3. Structured Programming: this module teaches learners introductory Python or R. Learners will be able to write short scripts and work in IPython Notebook, RStudio, or similar environments. This module will also teach key ideas—iteration, conditional statements, and modularity—that are essential for computational competence.
  4. Version Control and Data Sharing: this module teaches learners how to keep track of their code, data, and analyses in an open and reproducible way. The lessons will focus on data management strategies as well as provide an introduction to GitHub for version control of scripts, programs, analyses, etc. The real lessons here are about conducting open, reproducible research.

As the module descriptions suggest, our real aim isn't to teach Python, Git, or any other specific tool: it's to teach computational competence. We can't do this in the abstract: people won't show up for a hand-waving talk, and even if they do, they won't understand. If we show them how to solve a specific problem with a specific tool, though, we can then lead into a larger discussion of how scientists ought to develop, use, and curate software.

These workshops strive to show people how the pieces fit together: how to write a Python script that fits into a Unix pipeline, how to automate unit tests, etc. Doing this gives us a chance to reinforce ideas and also increases the odds of participants being able to apply what they've learned once the workshop is over.

Execution: Quality Instruction

We will aim for no more than 40 people per room at a workshop, so that every learner can receive personal attention when needed. Where possible, we will run two or more rooms side by side and use a pre-assessment questionnaire as a sorting hat to group learners by prior experience, which simplifies teaching and improves their experience.

All of the workshop instructors will have been trained and certified by Software Carpentry and will have had experience teaching this material prior to engaging in these particular workshops. Just as importantly, instructors will themselves be working scientists. By virtue of using these skills and concepts daily in their own research they are better able both to serve as role models and to deal with unanticipated questions or challenges based on personal experience.

Software Carpentry has a rich network of trained instructors, often recruited from past workshops. These instructors are volunteers who participate for a variety of reasons including sharpening their own teaching and computing skills, increasing diversity in the pipeline, and because it's fun.

As well as instructors, we will rely on local helpers to wander the room and answer questions during practicals. These helpers may be participants in previous workshops who are interested in becoming instructors, graduate students who've picked up some or all of this on their own, or members of the local open source community; where possible, we will aim to have at least one helper for every eight learners.

Increasing Diversity: Changing the Odds for Underrepresented Scientists

In order to increase the diversity of the study population, at least one workshop in each year will be aimed specifically at female students. Software Carpentry's first such workshop, held in Boston in June 2013, attracted 120 participants; its second is scheduled for Lawrence Berkeley National Laboratory in April 2014, and at least two more will be held by the time work on this project commences (one in the United States and one in Europe). This work will build on that experience and draw on the pool of instructors who have gained mentoring experience through those specific workshops.

Finally, we will organize workshops in years 2 through 5 specifically aimed at students from minority serving institutions. We are already in contact with the Computing Alliance for Hispanic-Serving Institutions (CAHSI) and with the Association of Public and Land-grant Universities' program for historically black colleges and universities (HBCUs). Software Carpentry is running its first workshop at an HBCU (Spelman College) in early 2014, and we expect to have significantly expanded these efforts by year two of this project.

Formative and Summative Assessment: Maximizing Learning and Impact

We will employ an expert in educational assessment full-time for five years to monitor and compare undergraduate participants in these software carpentry skills building workshops, participants in a subset of our regular (graduate-level) workshops, and non-participants (as a control population). As part of their work, this person will be responsible not only for collecting and analyzing data but also for refining and extending the methods and measures used to gauge impact. D-Lab will assist in locating and supporting this expert.

Assessment will build on previous work, focusing particularly, but not exclusively, on the following questions:

  1. Are students who receive this training more likely than their peers to develop new tools and practices and/or become involved in outreach and education activities (i.e., are they more likely to become creators and leaders)?
  2. Are students who receive this training more likely than their peers to incorporate open science and/or web-enabled science tools and practices into their work?
  3. Do outcomes differ between women and underrepresented minorities on one hand and non-underrepresented minorities and men on the other? If so, in what ways, and what steps are effective in correcting for these differences?
  4. In what ways does this training change students' outlook on the practice of science itself?
  5. Are students who receive this training more likely than their peers to choose computationally oriented research topics and/or careers? Are those who do not choose computationally oriented paths nevertheless more likely to incorporate the tools and practices mentioned above into their work?
  6. Are students who receive this training more likely to continue to graduate school than their peers?

This expert in educational assessment will explore ways in which our engagement with students changes the outlook and work practices of their peers and faculty supervisors (i.e., whether there is knowledge transfer sideways and upward) and the effectiveness of the community building and dissemination activities detailed in the next sections.

As with Software Carpentry's work to date, assessment will use both qualitative and quantitative techniques. On the qualitative side, we will conduct a series of interviews over the five-year period of the study to see how attitudes, aspirations, and activities change. Quantitatively, we will measure uptake of key tools such as version control as a proxy for adoption of related practices, as well as exploring more traditional measures of research success, such as progression to graduate school and publication/citation rates.

To track impact over time, we will conduct pre-workshop surveys and interviews roughly in the week before the workshops and the series of post-workshop surveys and interviews beginning approximately one month following the workshop. Follow on surveys and interviews will be conducted and the end of the students' lab appointment and annually thereafter (as applicable based on year of participation in the project).

Our findings, and any new methods or measures developed, will be shared with other researchers through publication in peer-reviewed journals and high-profile conferences.

Community Building: Supporting Computational Competence

We will employ one graduate student part-time at each site named in this proposal each year to provide technical support to workshop participants, and to act as an anchor for a Hacker Within-style grassroots group at that site. These community liaisons will not be study subjects but will help us stay in touch with students who are (a key requirement for any longitudinal study).

Separately, the Mozilla Science Lab will focus part of its ongoing community engagement efforts on the students who have taken part in our workshops during both the remainder of their undergraduate careers and afterward to help them become part of the broader open science community. This may include helping the students organize and run workshops of their own in subsequent years, connecting them with other open science projects, introducing them to potential graduate supervisors who understand and value their new skills and outlook, etc.

As a subordinate part of their work, the assessment researcher employed by this project will assess the effectiveness of the local graduate student organizers. In particular, they will explore whether seeding activity in this way leads to the formation of self-sustaining grassroots groups, and if so, what activities those groups develop on their own, how (and how effectively) they share discoveries with each other, the extent to which alumni of this program stay engaged with these groups, and whether the presence of these groups has a demonstrable impact on students' career paths in general and/or on their engagement with open and web-enabled science in particular.

While it will not be feasible to bring all of the students participating in a given year's workshops together physically, we will organize and run virtual conferences toward the end of their research term to give them an opportunity to present their work to one another, discuss what they have learned and build peer-to-peer connections. These conferences will also provide an opportunity to introduce participants to new forms of scientific "publishing", including blogging, the creation of screencasts and demonstration videos, and other methods that may not yet exist.

Curriculum Development and Dissemination: Expanding the Impact

We will employ one instructional designer part-time throughout the project to create new material and to improve existing material based on feedback from workshop participants and the assessment program. Here, "creating material" may include both designing and implementing new domain-specific learning modules and translating existing materials into new forms, such as video recordings of lectures or auto-graded quizzes for self-paced instruction. This work will be done in consultation with educators at participating institutions in order to encourage incorporation of those materials into existing curricula.

All of the materials produced by and for this project will be made freely available under the Creative Commons – Attribution (CC-BY) license. The instructional designer will work with the Mozilla Science Lab and affiliated groups to share these materials and the results of our studies of the program's impact, through science education journals, conferences, and other channels.

The dissemination of this project's curriculum has strong potential to be high. Current Software Carpentry materials are available as online lessons and in GitHub. Workshop materials will continue to be open access and flexible, thus they can be readily adopted by others. Adapting workshop materials is low cost and does not require special equipment. Workshop materials are structured such that they can scale to the size and application of interest to a particular group. Anyone using workshop materials can directly contribute changes and feedback, which both increases buy-in and improves the materials organically. The Software Carpentry infrastructure provides support in the form of materials and people. And finally, local chapters of The Hacker Within create a natural ecosystem of support for workshop participants, their peers, and faculty.

As a subordinate part of their work, the assessment researcher employed by this project will assess the extent to which curriculum developed during this program is taken up by other educators (particularly those who think of themselves as scientists first and computationalists second), and their perception of its utility. Mid-point results of this evaluation will be shared with the instructional designer in order to allow evidence-based improvement of the materials.

Project Management

Prof. Slaybaugh will be responsible for overall project management and reporting. The educational assessment expert hired by this project will report directly to her. Prof. Slaybaugh will be assisted by Dr. Huff, who will manage and coordinate the graduate student assistants at each site. Dr. Huff will also be responsible for organizing the workshops aimed at female students.

Profs. Barba, Brown, White, and P. Wilson will be responsible for coordinating workshops and for recruiting and supervising the graduate student assistant at their respective institutions. Prof. Teal will co-coordinate the workshops held at Michigan State University and will assist in developing workshop materials.

Dr. G. Wilson and the half-time instructional designer hired by this project will be responsible for preparation and publication of learning materials. Dr. G. Wilson will also provide instructional training for the graduate student assistants and other participants in the project on an ongoing basis and will be responsible for organizing the workshops aimed at students from minority serving institutions.

Workshop operations (such as finding instructors and arranging their travel) will be handled by Mozilla staff who are performing these duties for the Software Carpentry program more generally. These staff will be supervised by Ms. Thaney, who will also be responsible for connecting the other PIs and the graduate student assistants with other open and web-enabled science groups.

Related Work

Theoretical Positioning

Our theory of action is straightforward: if students are explicitly taught software skills in a way that makes them seem both useful and important, they are likely to begin using them in day-to-day work, which will create a positive feedback cycle leading them to acquire more (and more advanced) skills on their own. This positive feedback cycle will in turn result in the students being more likely to engage in open and web-enabled scientific activities that would otherwise have been unknown, incomprehensible, or out of reach.

Using the terminology of, our work is primarily design and development research. We plan to design and develop solutions related to student engagement and mastery of specific skills, drawing on existing evidence from Software Carpentry's graduate-level workshops and investigating their impact and effectiveness. Further, we further plan to design and iteratively develop interventions. We are ready to begin collecting data on the feasibility of implementing solutions in typical delivery settings.

Research

Studies of how scientists use computers and the web have found that most scientists learn what they know about developing software and using computers and the web in their research haphazardly and through word of mouth. In our experience, most training meant to address this issue:

  • does not target scientists' specific needs (e.g., is a general "Introduction to Computing" class shared with students majoring in other areas);
  • only covers the mechanics of programming in a particular language rather than giving a complete picture, including data management, web-enabled publishing, the "defense in depth" approach to correctness discussed in, or the other foundational skills laid out in; and/or
  • jumps to advanced topics such as parallel computing before scientists have mastered the foundations. Most research on scientific computing, such as, does the same.

On the other hand, studies of how people in general learn to program, and of how effective different approaches to teaching them are, have made significant strides in the past decade. In particular, our work is informed by the long-running research program of Guzdial et al. at Georgia Tech, who have found that a "media first" introduction to computing outperforms more conventional alternatives and that it is possible to assess the extent to which programming concepts, rather than merely the syntax of a particular programming language, have been mastered.

Others have demonstrated that peer instruction is a significantly better way to teach introductory programming than conventional classroom approaches. As discussed in the section below, we are already working to incorporate these evidence-based approaches into our teaching and will accelerate these efforts within the scope of this award.

Software Carpentry

Software Carpentry is the largest effort to date to address issues surrounding inadequate software carpentry skill training for students. Originally created as a training program at Los Alamos National Laboratory in the late 1990s, it is now part of the Mozilla Science Lab's efforts to help scientists take advantage of ways in which the web can change the practice of science today and invent new ways tomorrow. Over 100 certified volunteer instructors delivered two-day intensive workshops like those described earlier to more than 4200 people in 2013 alone.

Software Carpentry's curriculum and teaching practices have been refined via iterative design and are informed by current research on teaching and learning best practices. Its instructor-training program, which takes 2–4 hours/week of a trainee's time for 12–14 weeks (depending on scheduling interruptions), introduces participants to a variety of modern teaching techniques (e.g., peer instruction, active learning, and understanding by design), to concepts underlying these techniques (e.g., cognitive load theory), and to topic-specific work by computing education researchers (see,, and the first third of for overviews). One example of how they translate theory into practice is their insistence on live coding during teaching as a way of demonstrating and transferring authentic practice to learners.

Evidence of instructor training and experience is based on a Mozilla's Open Badge program. Further levels of expertise are obtained through helping at a workshop, being an instructor, being a lead instructor, and developing workshop materials. All of the participants in this proposal are certified instructors and have taught or will teach Software Carpentry workshops prior to the start of this project. For example, Dr. Huff has already taught at 9 workshops and Dr. G. Wilson has served as an instructor at 35.

Past Assessment

Software Carpentry has been assessing learning outcomes and retention since the beginning of its Sloan Foundation funding in January 2012. The first round of assessment included both qualitative and quantitative assessment by Dr. Jorge Aranda (then at the University of Victoria) and Prof. Julie Libarkin (Michigan State University).

Dr. Aranda surveyed and interviewed participants, observed a workshop, and analyzed screencasts of participants working through a programming assignment. The surveys and interviews were conducted both pre- and post-workshop. The survey asked questions regarding the software development habits of respondents, the tools they were familiar with, their level of knowledge of five core Software Carpentry topics (shell commands, Python, version control, SQL, and testing concepts), and their challenges in using scientific computing to answer their research questions.

According to both qualitative and quantitative data, Dr. Aranda found significant increases in participants' understanding and use of shell commands, version control tools, Python, and testing techniques. Perhaps more importantly, participants reported better proficiency with software tools; greater concern for issues of provenance and code quality; better strategies to approach software development; and new research questions that have become accessible thanks to an increase in participants' software development skills.

Students also took a quiz consisting of yes/no questions that were purposefully chosen so that only about half of them could be answered with the standard material in the workshop; the other half was not covered by workshop instruction. Additionally, a cross-cutting half could be answered with introductory familiarity to the topic in question, while the other half would represent more advanced levels of expertise.

The objectives for this were to assess whether participants not only learned the workshop materials but were exploring the topics in greater depth on their own, and to avoid ceiling effects in our survey. Quiz performance improved across the board (by ~30%) for all question categories.

Prof. Libarkin performed a more detailed assessment of participants in a workshop held at Michigan State University, which was attended remotely by students from the University of Texas at Austin. Prof. Libarkin also collected qualitative and quantitative data. Eighty-five percent of participants reported that they learned what they hoped to learn, 81% changed their computational understanding, and 96% said they would recommend the workshop to others.

An attempt to scale this up in 2013 was set back by personnel changes, but systematic follow-ups with past participants in workshops have now been resumed, and we expect to be able to present preliminary results by mid-2014.

Some excerpted survey questions from the discussed studies can be found in the full proposal, as can excerpted interview questions from Dr. Aranda's study. The studies mentioned here and the additional data that will become available within the next six months will serve as a starting point for this project's assessment expert.

The Hacker Within

The Hacker Within (THW) was founded by graduate students, including Dr. Huff and Prof. Slaybaugh, in nuclear engineering at the University of Wisconsin – Madison to provide a forum for sharing scientific computing skills and best practices with their peers. As THW matured as a student organization, it attracted students from many scientific disciplines and academic levels. THW conducted bi-weekly seminars and developed a series of short courses addressing the programming languages C++, Python, and Fortran; best practices such as version control and test-driven code development; and basic skills such as UNIX mobility. This curriculum was delivered primarily as interactive short workshops on campuses and during scientific conferences. Many previous founders of the Hacker Within have since become instructors with Software Carpentry, and a new generation of THW graduate students has begun to emerge in their place. In 2013, new branches of THW were initiated at the University of Melbourne and the University of California, Berkeley (under the direction of Dr. Huff and Prof. Slaybaugh).

Condo of Condos

The "Condo of Condos" consortium, led by Clemson University and including the University of Wisconsin – Madison and four other campuses during its pilot phase, has recently been recommended for funding by the National Science Foundation. The consortium's primary task is to build a network of advanced cyberinfrastructure research and education facilitators (ACI-REFs), with goals that include increasing the diversity of researchers using advanced cyber infrastructure on each campus and developing data science practitioners. The Software Carpentry workshops being designed under this proposal will serve those goals directly. As an investigator on both that project and this proposal, Prof. P. Wilson will engage the network of ACI-REFs to share this curriculum with both the initial consortium institutions and any institutions who are able to join in the future.

D-Lab

D-Lab, located at the University of California, Berkeley, has the mission to create cross-disciplinary resources for high-level training and support services for social science researchers campus-wide. It does this by adaptively building new forms of shared infrastructure, including consulting and workshops in research tools and methods, and fostering discovery and connections with the pools of expertise and offerings of Berkeley's departments and professional schools. As a new research unit at Berkeley, D-lab has been actively involved in teaching software tools. D-lab is engaged in evaluating students' experiences with those trainings and adapting subsequent trainings in response to their evaluations. D-lab is an enthusiastic supporter of campus efforts to broaden this approach to teaching.

D-lab will provide space, outreach, and assistance in locating and supporting the assessment position outlined in this proposal. Prof. Slaybaugh will engage with D-lab to capitalize on their expertise at the intersection assessment and scientific education.

Broader Impact

We believe this work will have significant impact in several areas beyond directly improving the computational science skills of workshop participants.

  1. Enhance economic competitiveness. Computing is no longer optional in any part of science: even scientists who don't think of themselves as doing computational work rely on computers to prepare papers, store data, and collaborate with colleagues. The better their computing skills are, the better prepared they will be to contribute to the research that underpins the nation's economic competitiveness.
  2. Improving STEM education for everyone, not just participants. By creating and validating high-quality open access teaching materials, and the methods used to deliver them, this project will enable improvement in STEM education for everyone, everywhere, not just for participating students and participating institutions.
  3. Improving STEM education tomorrow, not just today. As noted in the introduction of this proposal, most of today's efforts to transfer computational skills to STEM researchers and connect them with 21st Century innovations in how science is done are flying blind: there is effectively no feedback from long-term impact to instructional action. By creating and validating such a feedback loop—i.e., by showing scientists how to apply science to their teaching—this project will demonstrate how STEM education can be continuously improved.
  4. Improve participation in STEM by women and underrepresented minorities. The disproportionately low participation of women and some minority groups in STEM is well documented, as is the fact that computing is one of the least diverse fields within STEM. This second fact creates a vicious cycle: people with weaker computing skills may be less competitive in research than their peers, which reduces their participation in activities viewed as non-core, which in turn results in them having weaker skills. This project will strive to break this cycle by giving at-risk students an opportunity to "level up" in a supportive environment and by connecting them with mentors who can serve as role models.

Career Management Plan

The graduate students who are serving as mentors for the undergraduates at the different universities will each be paired with a local faculty mentor. The faculty mentor will meet regularly with the graduate student to discuss and problem solve any issues that the graduate student or undergraduates are having and to provide active mentoring on how to train students in computational approaches.

In addition to engaging with the graduate students on their mentoring of the undergraduates, the faculty mentors will also serve as mentors for computational aspects of the graduate students' research and careers. In many areas of science, computationally-minded students are located in labs where the PIs do not have strong computational backgrounds. This means that they do not have a mentor to teach them about good computational practice in research. In addition, they do not have someone with whom to discuss computational careers, thus limiting their exposure to career paths outside of academia. Because the faculty mentors will have strong computational backgrounds themselves, they can fill this void for computationally-minded students.

Dialogue & Discussion

You can review our commenting policy here.