Abstracts

POSTERS

The Guided Approach to Programming (GAP): Guiding Secondary School students in Problem Solving and Program Development.

Camille Berry
The University of the West Indies, Mona Campus
camille.berry02@uwimona.edu.jm

Abstract:

Problem Solving is an important skill that individuals need in order to function effectively and efficiently in the 21st century and beyond. Program Development or programming is the art of creating computer based solutions, using a language that can be interpreted by computers, which have helped humans to complete daily tasks. Problem solving and programming are therefore at the heart of innovative technologies and useful applications or real world solutions. In the Caribbean, problem solving and programming are taught at the secondary level as part of the Caribbean Secondary Examination Certificate (CSEC) Information Technology syllabus. The students have been struggling with these topics in the syllabus and have continued to perform unsatisfactorily in the aforementioned areas of the CSEC Examination (CSEC Report, 2014). In light of this, the teaching of problem solving and program development was investigated and a model was created to help the teachers guide the students in solving specific problems and subsequently in creating the computer programs based on the proposed solutions. The Guided Approach to Programming (GAP) model was designed on the basic principles of problem solving and was modified to include smaller steps that would help the students to think through all the possible elements of the solution. The process in designing this guide and the motivation behind the inclusion of the specified steps will be discussed in this presentation. Plans for future research will also be presented.

Thinking Ahead: Pre-registration

Neil C. C. Brown
King’s College London, UK

Abstract:

A recent attempt to replicate psychology studies found that many original effects did not replicate [1]. This so-called replication crisis prompted introspection about the causes, which may include p-hacking, adjusting hypotheses after results are known, the file drawer effect and other causes. Many of these bad research practices can be guarded against by following a core good practice: planning the data analysis before data collection, then making clear (and justifying) any post hoc adjustments. This good practice can be self-enforced and certified by pre-registering the study’s methods: recording them before data collection, for reviewers and readers to later examine.

The comprehensive form of pre-registration is registered reports, which involve the study being registered with a target journal. Reviewers then review the study plan and make an acceptance decision before data collection is carried out. This allows weak designs to be amended or rejected before effort is wasted on data collection, and eliminates the “file drawer” effect where negative results do not get published: the publication decision is no longer conditional on the results, only on the data collection following the pre-registered protocol. However, registered reports require a participating journal.

A simple alternative, which is immediately available, is to voluntarily pre-register. The Centre for Open Science have a website (http://osf.io/) where researchers can upload materials describing their method with a certified timestamp. This record can be used in publications to demonstrate that the analysis was pre-registered before data collection. Anonymous views of the material can be created for blind peer review.

This lightning talk and poster will explain pre-registration and try to increase interest from authors (and potentially journal editors) in computing education in adopting the practice, to help improve the quality of our research field.

[1] Estimating the reproducibility of psychological science, Open Science Collaboration, Science 349(6251), 2015.

A Bioinformatics Approach to Pattern Recognition in Open-Ended CSProgramming Assignments

Nati Ghatan
Sense Education

nati@sense.network, keren@sense.network

Abstract:

A central challenge in education is that of scalability — making education increasingly accessible and personalized in order to support students in fulfilling their potential in an ever-growing classroom, being either online or on campus. Especially in the field of Computer Science (CS), modern technological solutions attempt to partially alleviate the challenge by democratizing educational content, through MOOCs and other online content platforms. However, these solutions lack the essential component of educational feedback. Increasing access to content may even exacerbate the challenge by dramatically increasing the student-per-educator ratio, overwhelming the educators and making it nearly impossible to provide their students with high-quality personalized feedback.

We developed a machine learning solution that derives concepts from computational biology, genomics and bioinformatics in order to encode, discover and analyze patterns in non-biological datasets. By applying this technology to the field of CS, we automatically identify recurring response patterns in thousands of submissions for open-ended programming assignments. These patterns are then utilized to cluster the submissions, within minutes, into a small number of groups. Each cluster comprises a defined subset of submissions which represent a distinct solution archetype. Thus, an educator may quickly examine and provide tailored educational feedback by referring to only several clusters and patterns, rather than manually assessing individual submissions. Furthermore, as clusters are defined by the unique combination of their response patterns, they carry information about the internal logic of the submissions. This allows the educator to examine intricate components in the students’ approach to a problem, as opposed to existing methods (e.g. output correctness testing) which provide quick batch assessments at the expense of quality by focusing on narrow pedagogical aspects.

Designing Hackathons for Equity: the literature’s best advice

Caroline D. Hardin
PhD Student
Computer Science Education
UW Madison

Abstract:

Hackathons are an increasingly popular platform for computer science collaborative learning. As much as the literature is overwhelmingly positive about the potential of hackathons, there is agreement that the limited demographics of the attendees is problematic. While the literature also has a lot to say about how to design hackathons which might have less embarrassing demographics in attendance, these ideas have not yet been organized into a single source. In addition, they have not been critically analyzed for effectiveness or alignment with the goals of orienting participants to authentic CS practices. What interventions should be recommended to be adopted at scale, or where there are gaps which need to be addressed by the design of new interventions, therefore, is unknown. This work examines over 50 prominently cited papers, popular press articles, blog posts and other sources which included ideas or discussion about how to increase equity (or at least diversity) at hackathons. All the design ideas are organized according to the type of commitment they require on the part of the organizers, which roughly correlates to the likelihood of being effective. The ideas are divided into four categories: 1) appear welcoming 2) invest time in changing the experience 3) invest money in changing the experience 4) build relationships which share power. Each of these categories is discussed, followed by a list of prominent examples.The literature has many pieces of advice for increasing equity in CS through hackathons which is compatible with authentic practices in CS. The goal of this study is to provide a guide on how to understand this tension, followed by a comprehensive list of advice present in literature for those who are interested in organizing better hackathons for all.

Computational Thinking for All Pre-Service Teachers

Jennifer Rosato
Jill Long
Chery Lucarelli
College of St. Scholastica
Heather Benedict
Heather Benedict, LLC

Abstract:

As more schools in the United States offer computer science (CS), through an Hour of Code event or courses like CS Principles, teachers are being asked to teach an unfamiliar subject – CS. In order for the CSforAll movement to be truly equitable and sustainable, reaching all schools and all students, all pre-service teachers must learn CS and it’s connections to their licensure area (the subjects and grades that they will be teaching) [1, 2].

The College of St. Scholastica’s School of Education is piloting the integration of computational thinking (CT) and CS throughout its teacher preparation programs. Each program provides opportunities to learn and engage at a level appropriate to the various licensures. Teacher candidates complete a CT module introducing basic concepts and making connections to equity, pedagogy, and the broader CS education community (curriculum, standards, etc.)

This module is then reinforced and expanded on in the methods and field experience courses. Education faculty offering elementary and secondary methods courses have developed lessons integrating CT/CS and methods subjects (e.g. science, math, visual arts) with support from CS curriculum coaches. Teacher candidates will then have a field experience where they practice teaching CT/CS to K12 students, mentored by local cooperating teachers and education faculty supervisors.

This poster will share results from the first two years of the project, including education faculty development, the computational thinking module and integrated CT/CS lessons, and initial work with partner schools and cooperating teachers to prepare for field experiences. Data on coaching education faculty and lessons learned from this pilot project will be shared as well. Data from the first cohorts of students will be available after the fall semester.

Intellectual Need in Programming Activities

Vinesh Kannan
Mimir; Illinois Institute of Technology
v@hawk.iit.edu

Abstract:

Motivations for learning programming commonly fall under an economic need, such as improving employment prospects, or a social need, such as passing a required computer science class. But these needs alone do not motivate every valuable topic in computing education.

Fuller, Rabin, and Harel describe five categories of intellectual need which, while based on research in mathematics classrooms, explicitly highlight computing concepts such as runtime complexity [1]. In the United States, the expansion of K-12 computer science education leads many mathematics instructors to teach computing. Professional development for computing education can empower such instructors with a framework they may already be confident in.

The intellectual need framework also relates to two of the four pedagogical principles Kim and Ko suggest for improving online coding tutorials: connecting to learners’ prior knowledge and encouraging meta-cognitive learning [2]. Learners sometimes describe programming problems as contrived or pointless. Such perspectives align with Fuller, Rabin, and Harel’s description of problem-free activity [1]. Applying the intellectual need framework to computing curricula may help learners engage with content they might otherwise consider obscure.

This presentation aims to solicit feedback on how intellectual need might differ between computing and mathematics education as well as how intellectual need might influence learning transfer. The lightning talk and poster demonstrate a suite of activities designed as open educational resources (OER) that emphasize each category of intellectual need. 52 novice programmers completed the activities in a four-week online course during summer 2018. Instructors analyzed learners’ daily reflections, programming task scores, and transcripts of online tutoring conversations to identify instances of intellectual need and problem-free activity.

[1] Evan Fuller, Jeffrey Rabin, and Guershon Harel. 2011. Intellectual Need and Problem-Free Activity in the Mathematics Classroom.

[2] Ada S. Kim and Andrew J. Ko. 2017. A Pedagogical Analysis of
Online Coding Tutorials.

An Artificial Divide? Skills Development in Boot Camps vs. Academic CS Preparation at the University

Louise Ann Lyon
ETR
LouAnn.Lyon@ETR.org

Abstract:

The rise in the U.S. of so-called “coding boot camps” offering postsecondary training in computer science-related topics highlights an ongoing debate about the nature of higher education, the categorization of learning settings into vocational or academic locations, and the responsibility of these organization to their students. Our work reports on qualitative research interviews with boot camp organizers, computer science (CS) university faculty, and students in each setting at two timepoints (once as students and again after six months on the job) to illustrate case studies along the workforce preparation-to-academic preparation continuum and to contextualize the type of students in both settings, the learning in each environment, and job outcomes for students from different settings. We place four sites along the continuum-two boot camps and two universities-according to their stated goals, pedagogical approaches, and views on the purpose of their institution. Findings show that some boot camp organizers and university faculty are dedicated to the idea that preparing students to enter the workforce is a social justice issue, while others are committed to providing “foundational” CS education assumed to create a basis for later skills learning. Findings from student data give hints as to why some students choose boot camps over university CS programs as well as learner perspectives on job preparation and preparation for future learning. Discoveries from this study are designed to spark a conversation with ICER participants about what constitutes student “success” and varying viewpoints on the “value” of different settings.

Women’s Informal Coding Groups as Incubators: Legitimate Peripheral Participation within a Subset Community of Practice

Louise Ann Lyon
ETR
LouAnn.Lyon@ETR.org

Abstract:

Female-focused, grassroots communities of practice purporting to help women learn to code are popping up in a variety of settings, indicating the motivation on the part of the participants to avoid male-dominated settings while learning. This study used an ethnographic approach to investigate an emergent coaching and learning community designed to help women move from Salesforce CRM (Customer Relationship Management) platform administrators to developers through learning to code. This all-volunteer group has built an Apex language–similar to Java–curriculum and continues to form cohorts where novices meet virtually for 10 weeks with coaches and other learners to learn coding, share code, and ask questions based on weekly individual practice. Using Lave and Wenger’s legitimate peripheral participation framework, we investigated if and how interactions with others in this community bolstered increased identification as a “developer” among the participants as well as how interactions with both coaches and other learners enhanced the learning of content. Findings indicate that a women-focused community of practice can be an incubator for later full participation in and identification with the larger software developer community of practice. In addition, findings illustrate that those who value the feminine nature of the group note two advantages to their learning: increased comfort in asking “stupid” questions and abundant verbal encouragement. Confidence gained by participants through supportive interactions and foundational knowledge lasted beyond the duration of group as women re-engaged with the larger Salesforce developer community. These findings have implications for the CS education community in emphasizing the importance of creating safe spaces to ask novice questions as well as the importance of verbal reassurance for many women learning to code.

Assessing the unseen: Roles of Confidentiality and Trust in software engineering work-based learning programmes.

Joseph Maguire – School of Computing Science
Steve Draper – School of Psychology
Quintin Cutts – School of Computing Science
University of Glasgow
joseph.maguire@glasgow.ac.uk, s.draper@psy.gla.ac.uk, quintin.cutts@glasgow.co.uk
www.dcs.gla.ac.uk/~joseph, www.psy.gla.ac.uk/~steve, www.dcs.gla.ac.uk/~quintin

Abstract:

A typical academic degree focused on software engineering has little practical relationship with the industry it is named for, other than the occasional placement or internship. Unlike other professions such as medicine, dentistry and veterinary sciences, candidates do not need to participate in significant professional practice to earn their degree. Indeed, if we consider a traditional academic software engineering student they probably have far more experience constructing shiny new ‘green-field’ systems, than maintaining the old ‘brown-field’ systems found in industry, and generating most professional work. Consequently, there is growing enthusiasm for work-based learning programmes that provide an opportunity for candidates to cement abstract academic theory in concrete personal experience. Work-based learning software engineering students earn their degree by combining theory with actual practice in a professional environment.

Nevertheless, the intangible outcomes for much of software engineering has led to an industry obsessed with confidentiality, driven by concerns of employees smuggling source code to competitors or regulators. This obsession potentially presents a barrier to work-based learning schemes as employers prevent outsiders, even close higher education partners, from observing the systems and the source code that learners are working on. Learners may have the opportunity for concrete personal experience, but educators are barred from observing any such experience.

However, confidentiality agreements may not necessarily present barriers to assessment, but instead provide an opportunity to assess comprehension and transferable skills by requiring abstract descriptions and reports. This is the converse to the problem in some programming courses, where students submit code without demonstrating that they understand it and can discuss it in terms of the concepts taught.

In this talk and accompanying poster we explore some models for software engineering work-based learning programmes that have the potential to maintain confidentiality while assessing learners’ comprehension and ability. We invite discussion and criticism from conference attendees of the presented models, and are interested in potential partners for future collaboration.

Robots for Novices via Computational Design

Devon James Merrill (UC San Diego)
Steven Swanson (UC San Diego)

Abstract:

Hands on experience is extremely valuable for novices. However, embedded design and electronics labs require a notoriously large amount of instructional staff time. We have developed a web-based robot design tools that allows novices to quickly create custom robots in a matter of minutes with little assistance.

This tool has enabled us to develop a course for first quarter, freshmen computer science students with no embedded design experience. In this course, students build and program custom robots. Each robot includes a custom circuit board, computationally designed firmware and API, and computationally designed assembly instructions. The tool’s automated and end-to-end nature enables the course to be taught with minimal instructional staff.

The tool allows student to specify a circuit board shape that forms the mechanical substrate of their robot. They drag-and-drop functional components onto the design, such as motorized wheels, grippers, bump sensors, lights, and buttons. Students do not need to specify any electrical details about their design. The tool automatically fills in electrical connections between the components, support components such as motor drivers or resistors, and pin assignments. The tool automatically creates, processes, and assembles all the needed manufacturing files, firmware, bill of materials, and assembly instructions.

We are currently collecting data on the effect of early hands-on electronics and embedded programming experience with regards to program retention rates and attitudes towards computer science, especially in under-represented groups. We are currently running user studies with our robot design tool as a domain for invention-based learning.

Engaging Students in CS with TurtleStitch

Briana B. Morrison
University of Nebraska Omaha
bbmorrison@unomaha.edu

Abstract:

As part of a summer Research Experience for Teachers (RET), two high school teachers designed an introduction to programming unit using TurtleStitch (turtlestitch.org) for high school students. TurtleStitch is a block-based programming environment based on the educational programming language Snap!. Turtlestitch uses Snap!’s “pen module” which it interprets as a needle and transforms its output into widely-used embroidery file formats. Students can then transfer the pattern to an embroidery machine and see their pattern “come to life”.

We designed a study to determine the effect of using digital embroidery as a medium for introducing students to programming. The pilot study was conducted at two Nebraska high schools, one an all-girls private high school in the Web Design course (8 students) and a public high school in the Introduction to Programming course (17 students). In the quasi-experimental mixed methods study, demographic information, pre and post attitudinal information was collected as well as daily journal entries during the six days of lessons. Survey results for student aptitude and interest for sewing/embroidery as well as programming all increased due to the intervention. Student aptitude for both sewing/embroidery and programming being statistically significant (p < 0.05).

The qualitative analysis supports the statistical findings, indicating that student aptitude for both sewing and programming increased during the intervention. Students enjoyed creating their own designs and then seeing the tangible results of their labor in the form of an embroidered T-shirt or hat. Students found creating a function to be the most difficult part of the assignment and incorporating loops was found to be the most rewarding.

For future work we plan to expand the number of schools and students using the lesson plan with more detailed quantitative and qualitative analysis.

Towards Generating Concept Inventories Using Student Contributions and Machine Learning

Sam Saarinen
Kathi Fisler
Shriram Krishnamurthi
Preston Tunnell Wilson

Concept Inventories (CIs) can be used as assessments of student understanding of a particular topic, but are challenging, expensive, and time-consuming to produce. We introduce a student-sourcing activity facilitated by machine learning that can generate question-answer-reasoning tuples akin to those on concept inventories. A tool implementing this activity has been deployed. In a study involving questions about arrays in Java which we compare to questions in an expert-constructed CI, this novel method: generates many of the same kinds of questions as the expert-constructed CI; produces informative questions not in the expert-constructed CI; and enables efficient validation of distractors (wrong answers) with associated misconceptions. These generated Quasi-Concept Inventory (QCI) questions can be leveraged to produce better CIs with less expert labor. This process also raises the possibility of running reproduction studies relatively easily.

Differences in Reporting of Outreach Research Data between CS Education and STEM Education

Adrienne Decker, Rochester Institute of Technology, Adrienne.Decker@rit.edu
Monica M. McGill, Knox College, mmmcgill@knox.edu

Abstract:

In previous systematic literature reviews of computing education [1, 2], we identified key shortcomings in the reporting of data and information for research studies concerning pre-college computing activities (sometimes called outreach). In the context of a larger project to provide a repository and resources for those interested in pre-college computing activities, a more thorough extraction of data was undertaken [4], the results of which are currently available through https://csedresearch.org.

As an additional reference point, we sought to compare computing education work in this area with other STEM fields. As such, a systematic literature review [3] of 21 STEM education journals and conference venues was conducted spanning the years 2014-2016 inclusive. This literature review identified 162 candidate articles to analyze for comparison data. The same variables used in the computing education literature review [4] were extracted from these STEM articles.

This work presents the initial findings in the comparison of the outreach literature of computing education and STEM education, identifying points of similarity as well as points of difference between the frequency of reporting and the style of reporting of the identified variables across the two sets of articles.

[1] Decker A., McGill, M.M., Settle, A. (2016) “Towards a Common Framework for Evaluating Computing Outreach Activities”, Proceedings of SIGCSE 2016, March 2-5, 2016, Memphis, TN, pp. 627632.

[2] Decker, A., McGill, M. (2017) “Pre-College Computing Outreach Research: Towards Improving the Practice”, Proceedings of SIGCSE 2017 March 8-11, 2017, Seattle, WA, pp. 153-158.

[3] Khan, K.S., Kunz, R., Kleijnen, J., Antes, G. 2003. Five steps to conducting a systematic review. J R Soc Med. 96, 3, 118–121. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC539417/

[4] McGill, M., Decker, A., Abbott, Z. (2018) “Improving Research and Experience Reports of PreCollege Computing Activities: A Gap Analysis”, Proceedings of SIGCSE 2018, February 21-24, 2018, Baltimore, MD, pp. 964-969.

LIGHTNING TALKS

Even Young Kids Can Learn CS with Robotics

Mordechai (Moti) Ben-Ari
Department of Science Teaching
Weizmann Institute of Science
moti.ben.ari@gmail.com

Abstract:

We taught CS and robotics in four second-grade classrooms (ages 7-8) with about 30 students in each class. The goals of this research project were:

1. To investigate learning CS with robotics in a normal classroom, not in a voluntary extracurricular activity.
2. To confirm or reject the Jourdain effect: Does successful performance of a task indicate understanding of a concept?
3. To characterize what can be learned by students of this age group.

We used the Thymio robot (https://www.thymio.org/), which has a low floor and a high ceiling. The VPL programming environment for the robot uses graphic blocks to construct event handlers.

Observations showed a very high level of engagement.

We developed questionnaires based on a new taxonomy of levels of understanding. The questionnaires used graphics and video clips to overcome the relatively low reading and writing abilities of such young children.

The analysis of questionnaires showed that for the majority of the students, the Jourdain effect did not occur: The students were able to understand the semantics of the individual blocks and to construct meaningful event handlers. However, the students were not able to plan and implement programs (Soloway and Spohrer),

The purposes of this talk are:

1. To encourage future research on characterizing what CS concepts can be learned by different age groups.
2. To encourage work on developing CS-specific taxonomies of learning and instruments for computing educational research in elementary schools.

M. Friebroon Yesharim, M. Ben-Ari. Teaching Robotics Concepts to Elementary School Children. 8th International Conference on Robotics in Education, Sofia, Bulgaria, 2017, 26–28.

Thinking Ahead: Pre-registration

Neil C. C. Brown
King’s College London, UK

Abstract:

A recent attempt to replicate psychology studies found that many original effects did not replicate [1]. This so-called replication crisis prompted introspection about the causes, which may include p-hacking, adjusting hypotheses after results are known, the file drawer effect and other causes. Many of these bad research practices can be guarded against by following a core good practice: planning the data analysis before data collection, then making clear (and justifying) any post hoc adjustments. This good practice can be self-enforced and certified by pre-registering the study’s methods: recording them before data collection, for reviewers and readers to later examine.

The comprehensive form of pre-registration is registered reports, which involve the study being registered with a target journal. Reviewers then review the study plan and make an acceptance decision before data collection is carried out. This allows weak designs to be amended or rejected before effort is wasted on data collection, and eliminates the “file drawer” effect where negative results do not get published: the publication decision is no longer conditional on the results, only on the data collection following the pre-registered protocol. However, registered reports require a participating journal.

A simple alternative, which is immediately available, is to voluntarily pre-register. The Centre for Open Science have a website (http://osf.io/) where researchers can upload materials describing their method with a certified timestamp. This record can be used in publications to demonstrate that the analysis was pre-registered before data collection. Anonymous views of the material can be created for blind peer review.

This lightning talk and poster will explain pre-registration and try to increase interest from authors (and potentially journal editors) in computing education in adopting the practice, to help improve the quality of our research field.

[1] Estimating the reproducibility of psychological science, Open Science Collaboration, Science 349(6251), 2015.

A Bioinformatics Approach to Pattern Recognition in Open-Ended CS Programming Assignments

Nati Ghatan
Sense Education
nati@sense.network, keren@sense.network

Abstract:

A central challenge in education is that of scalability — making education increasingly accessible and personalized in order to support students in fulfilling their potential in an ever-growing classroom, being either online or on campus. Especially in the field of Computer Science (CS), modern technological solutions attempt to partially alleviate the challenge by democratizing educational content, through MOOCs and other online content platforms. However, these solutions lack the essential component of educational feedback. Increasing access to content may even exacerbate the challenge by dramatically increasing the student-per-educator ratio, overwhelming the educators and making it nearly impossible to provide their students with high-quality personalized feedback.

We developed a machine learning solution that derives concepts from computational biology, genomics and bioinformatics in order to encode, discover and analyze patterns in non-biological datasets. By applying this technology to the field of CS, we automatically identify recurring response patterns in thousands of submissions for open-ended programming assignments. These patterns are then utilized to cluster the submissions, within minutes, into a small number of groups. Each cluster comprises a defined subset of submissions which represent a distinct solution archetype. Thus, an educator may quickly examine and provide tailored educational feedback by referring to only several clusters and patterns, rather than manually assessing individual submissions. Furthermore, as clusters are defined by the unique combination of their response patterns, they carry information about the internal logic of the submissions. This allows the educator to examine intricate components in the students’ approach to a problem, as opposed to existing methods (e.g. output correctness testing) which provide quick batch assessments at the expense of quality by focusing on narrow pedagogical aspects.

Teaching Exponential Organization: The Case of Computer Science Education

Orit Hazzan
Gadi Har-shai
Rotem Bennet
Ruti Donag
Technion – Israel Institute of Technology
oritha@ed.technion.ac.il, gadi.harshai@gmail.com, rotem.bennet@gmail.com, rutid@technion.ac.il

Abstract:

Exponential organizations (ExOs) are ones whose “impact (or output) is disproportionally large – at least 10x larger – compared to its peers because of the use of new organizational techniques that leverage exponential technologies.” [1] We are all familiar with such organizations (e.g., Google, TED, Uber, Airbnb, Waze, and many more). ExOs are characterized by 10 practices that contribute and foster their exponential growth (5 practices are internal to the organization and 5 – are external).

In the lightning talk we describe how we teach the concept of ExO in a course about soft skills taught to 50 undergraduate Computer Science students at the Technion – Israel Institute of Technology.

We found it relevant to expose undergraduate Computer Science students to the concept of ExO in general and its practices in particular for two main reasons. First, several practices are based on basic topics in Computer Science, such as algorithms (e.g., big data and machine learning), and software development methods. Second, it is reasonable to assume that many of the students will work in an ExO in the future.

A course on soft skills was selected for teaching the concept of ExO since the implementation of several ExO practices require the expression of soft skills, such as teamwork, communication, agility, and problem solving under constraints. Specifically, the concept of ExO was introduced and practiced in a hackathon in which the students developed innovative technological products for people with disabilities, while practicing working with real customers from the Technion Social Hub: https://socialhub.technion.ac.il/english-page/.

[1] Salim Ismail, Michael S Malone, Yuri van Geest. 2014. Exponential Organizations: Why new organizations are ten times better, faster, and cheaper than yours (and what to do about it), Diversion Books.

Women’s Informal Coding Groups as Incubators: Legitimate Peripheral Participation within a Subset Community of Practice

Louise Ann Lyon
ETR
LouAnn.Lyon@ETR.org

Abstract:

Female-focused, grassroots communities of practice purporting to help women learn to code are popping up in a variety of settings, indicating the motivation on the part of the participants to avoid male-dominated settings while learning. This study used an ethnographic approach to investigate an emergent coaching and learning community designed to help women move from Salesforce CRM (Customer Relationship Management) platform administrators to developers through learning to code. This all-volunteer group has built an Apex language–similar to Java–curriculum and continues to form cohorts where novices meet virtually for 10 weeks with coaches and other learners to learn coding, share code, and ask questions based on weekly individual practice. Using Lave and Wenger’s legitimate peripheral participation framework, we investigated if and how interactions with others in this community bolstered increased identification as a “developer” among the participants as well as how interactions with both coaches and other learners enhanced the learning of content. Findings indicate that a women-focused community of practice can be an incubator for later full participation in and identification with the larger software developer community of practice. In addition, findings illustrate that those who value the feminine nature of the group note two advantages to their learning: increased comfort in asking “stupid” questions and abundant verbal encouragement. Confidence gained by participants through supportive interactions and foundational knowledge lasted beyond the duration of group as women re-engaged with the larger Salesforce developer community. These findings have implications for the CS education community in emphasizing the importance of creating safe spaces to ask novice questions as well as the importance of verbal reassurance for many women learning to code.

Assessing the unseen: Roles of Confidentiality and Trust in software engineering work-based learning programmes.

Joseph Maguire – School of Computing Science
Steve Draper – School of Psychology
Quintin Cutts – School of Computing Science
University of Glasgow
joseph.maguire@glasgow.ac.uk, s.draper@psy.gla.ac.uk, quintin.cutts@glasgow.co.uk
www.dcs.gla.ac.uk/~joseph, www.psy.gla.ac.uk/~steve, www.dcs.gla.ac.uk/~quintin

Abstract:

A typical academic degree focused on software engineering has little practical relationship with the industry it is named for, other than the occasional placement or internship. Unlike other professions such as medicine, dentistry and veterinary sciences, candidates do not need to participate in significant professional practice to earn their degree. Indeed, if we consider a traditional academic software engineering student they probably have far more experience constructing shiny new ‘green-field’ systems, than maintaining the old ‘brown-field’ systems found in industry, and generating most professional work. Consequently, there is growing enthusiasm for work-based learning programmes that provide an opportunity for candidates to cement abstract academic theory in concrete personal experience. Work-based learning software engineering students earn their degree by combining theory with actual practice in a professional environment.

Nevertheless, the intangible outcomes for much of software engineering has led to an industry obsessed with confidentiality, driven by concerns of employees smuggling source code to competitors or regulators. This obsession potentially presents a barrier to work-based learning schemes as employers prevent outsiders, even close higher education partners, from observing the systems and the source code that learners are working on. Learners may have the opportunity for concrete personal experience, but educators are barred from observing any such experience.

However, confidentiality agreements may not necessarily present barriers to assessment, but instead provide an opportunity to assess comprehension and transferable skills by requiring abstract descriptions and reports. This is the converse to the problem in some programming courses, where students submit code without demonstrating that they understand it and can discuss it in terms of the concepts taught.

In this talk and accompanying poster we explore some models for software engineering work-based learning programmes that have the potential to maintain confidentiality while assessing learners’ comprehension and ability. We invite discussion and criticism from conference attendees of the presented models, and are interested in potential partners for future collaboration.

Robots for Novices via Computational Design

Devon James Merrill (UC San Diego)
Steven Swanson (UC San Diego)

Abstract:

Hands on experience is extremely valuable for novices. However, embedded design and electronics labs require a notoriously large amount of instructional staff time. We have developed a web-based robot design tools that allows novices to quickly create custom robots in a matter of minutes with little assistance.

This tool has enabled us to develop a course for first quarter, freshmen computer science students with no embedded design experience. In this course, students build and program custom robots. Each robot includes a custom circuit board, computationally designed firmware and API, and computationally designed assembly instructions. The tool’s automated and end-to-end nature enables the course to be taught with minimal instructional staff.

The tool allows student to specify a circuit board shape that forms the mechanical substrate of their robot. They drag-and-drop functional components onto the design, such as motorized wheels, grippers, bump sensors, lights, and buttons. Students do not need to specify any electrical details about their design. The tool automatically fills in electrical connections between the components, support components such as motor drivers or resistors, and pin assignments. The tool automatically creates, processes, and assembles all the needed manufacturing files, firmware, bill of materials, and assembly instructions.

We are currently collecting data on the effect of early hands-on electronics and embedded programming experience with regards to program retention rates and attitudes towards computer science, especially in under-represented groups. We are currently running user studies with our robot design tool as a domain for invention-based learning.

What Are Computer Science Educators Interested In? The Case of SIGCSE Conferences

Noa Ragonis – Beit Berl College & Technion
Orit Hazzan – Technion
Israel Institute of Technology
noarag@beitberl.ac.il, oritha@technion.ac.il

Abstract:

One of the challenges facing CS educators is coping with the rapid changes taking place in the discipline of CS and in students’ interests and skills. In order to prepare future computer scientists and software engineers for the largely unknown work market, lecturers and teachers require to cope with new subject matter contents, to prepare new teaching materials, and to diverse their teaching methods including the integration of up-to-date learning environments. Based on this recognition, we asked: What changes have taken place in CS education over the past decade, as reflected in the professional content of CSE conferences and journals. We started our investigation in relation to the SIGCSE Technical Symposium, since it is the largest venue that provides a forum for educators to discuss issues related to the development, implementation, and evaluation of computing study programs, curricula, courses, and pedagogy. For example, with respect to change in contents, the “Big Data” concept appeared first in SIGCSE 2011, and with respect to pedagogical changes, the “MOOCs” approach appeared first at SIGCSE 2012.

Based on the investigation of the last SIGCSE conference proceedings, we identified main topics, themes and trends in the following four presentation formats: papers, panels, special sessions, and workshops. We derived five content categories from the different presentations: teaching methods, curricula, CS education research, recruitment and retention, and educators. In this lightening talk, we present the frequencies of the topics presented in these conferences, which expresses changes and new foci. The investigation of the proceedings of the SIGCSE conferences is the first step of a wider investigation.

Towards Generating Concept Inventories Using Student Contributions and Machine Learning

Sam Saarinen
Kathi Fisler
Shriram Krishnamurthi
Preston Tunnell Wilson

Abstract:

Concept Inventories (CIs) can be used as assessments of student understanding of a particular topic, but are challenging, expensive, and time-consuming to produce. We introduce a student-sourcing activity facilitated by machine learning that can generate question-answer-reasoning tuples akin to those on concept inventories. A tool implementing this activity has been deployed. In a study involving questions about arrays in Java which we compare to questions in an expert-constructed CI, this novel method: generates many of the same kinds of questions as the expert-constructed CI; produces informative questions not in the expert-constructed CI; and enables efficient validation of distractors (wrong answers) with associated misconceptions. These generated Quasi-Concept Inventory (QCI) questions can be leveraged to produce better CIs with less expert labor. This process also raises the possibility of running reproduction studies relatively easily.