THE SKY IS NOT THE LIMIT: Multitasking Across GitHub Projects - PowerPoint PPT Presentation

THE SKY IS NOT THE LIMIT: Multitasking Across GitHub Projects Octocat, here and elsewhere, by GitHub https://octodex.github.com Bogdan Vasilescu (@b_vasilescu) Kelly Blincoe (@KellyBlincoe) Qi Xuan Casey Casalnuovo Dana Damian Prem Devanbu Vladimir Filkov 1247280 1414172

Multitasking is common #icsenumber

Software developers multitask too GitHub developer (25 Nov 2013 — 18 May 2014) EXAMPLE: #Projects 0 1 3 5 8 Mon Tue Wed Thu Fri Sat Sun Nov Dec Jan Feb Mar Apr

Software developers multitask too GitHub developer (25 Nov 2013 — 18 May 2014) EXAMPLE: #Projects 0 1 3 5 8 Mon Tue Wed Thu Fri Sat Sun Nov Dec Jan Feb Mar Apr

Software developers multitask too GitHub developer (25 Nov 2013 — 18 May 2014) EXAMPLE: #Projects 0 1 3 5 8 Mon Tue Wed Thu Fri Sat Sun Nov Dec Jan Feb Mar Apr

Software developers multitask too GitHub developer (25 Nov 2013 — 18 May 2014) EXAMPLE: #Projects 0 1 3 5 8 Mon Tue Wed Thu Fri Sat Sun Nov Dec Jan Feb Mar Apr

Software developers multitask too EXAMPLE: GitHub developer (25 Nov 2013 — 18 May 2014) #Projects 5 8 Mon Tue Wed Thu Fri Sat Sun Nov Dec Jan Feb Mar Apr

Software developers multitask too EXAMPLE: GitHub developer (25 Nov 2013 — 18 May 2014) #Projects 5 8 Mon Tue Wed Thu Fri Sat Sun Nov Dec Jan Feb Mar Apr WHY? ‣ Request from other dev’s / management

Software developers multitask too EXAMPLE: GitHub developer (25 Nov 2013 — 18 May 2014) #Projects 5 8 Mon Tue Wed Thu Fri Sat Sun Nov Dec Jan Feb Mar Apr WHY? ‣ Request from other dev’s / management ‣ Dependencies

Software developers multitask too EXAMPLE: GitHub developer (25 Nov 2013 — 18 May 2014) #Projects 5 8 Mon Tue Wed Thu Fri Sat Sun Nov Dec Jan Feb Mar Apr WHY? ‣ Request from other ‣ Being “stuck” dev’s / management ‣ Downtime ‣ Dependencies

Software developers multitask too EXAMPLE: GitHub developer (25 Nov 2013 — 18 May 2014) #Projects 5 8 Mon Tue Wed Thu Fri Sat Sun Nov Dec Jan Feb Mar Apr WHY? ‣ Request from other ‣ Being “stuck” ‣ Personal interest dev’s / management ‣ Downtime ‣ Dependencies

Software developers multitask too EXAMPLE: GitHub developer (25 Nov 2013 — 18 May 2014) #Projects 5 8 Mon Tue Wed Thu Fri Sat Sun Nov Dec Jan Feb Mar Apr WHY? ‣ Request from other ‣ Being “stuck” ‣ Personal interest dev’s / management ‣ Signaling ‣ Downtime ‣ Dependencies

Theory: How does multitasking affect performance? PROS CONS

Theory: How does multitasking affect performance? PROS CONS ‣ Fill downtime Switch focus between projects to utilize time more efficiently (Adler and Benbunan-Fich, 2012)

Theory: How does multitasking affect performance? PROS CONS ‣ Fill downtime Switch focus between projects to utilize time more efficiently (Adler and Benbunan-Fich, 2012) ‣ Cross-fertilisation Easier to work on other projects if knowledge is transferrable (Lindbeck and Snower, 2000)

Theory: How does multitasking affect performance? PROS CONS ‣ Fill downtime ‣ Cognitive switching cost Switch focus between Depends on interruption projects to utilize time duration, complexity, more efficiently moment (Adler and Benbunan-Fich, (Altmann and Trafton, 2002) 2012) (Borst, Taatgen, van Rijn, 2015) ‣ Cross-fertilisation Easier to work on other projects if knowledge is transferrable (Lindbeck and Snower, 2000)

Theory: How does multitasking affect performance? PROS CONS ‣ Fill downtime ‣ Cognitive switching cost Switch focus between Depends on interruption projects to utilize time duration, complexity, more efficiently moment (Adler and Benbunan-Fich, (Altmann and Trafton, 2002) 2012) (Borst, Taatgen, van Rijn, 2015) ‣ Cross-fertilisation ‣ “Project overload” Easier to work on other Mental congestion when projects if knowledge is too much multitasking transferrable (Zika-Viktorsson, Sundstrom, Engwall, 2006) (Lindbeck and Snower, 2000)

Theory: How does multitasking affect performance? PROS CONS ‣ Fill downtime ‣ Cognitive switching cost Switch focus between Depends on interruption projects to utilize time duration, complexity, In theory: more efficiently moment (Adler and Benbunan-Fich, (Altmann and Trafton, 2002) Productivity 2012) (Borst, Taatgen, van Rijn, 2015) Amount of multitasking ‣ Cross-fertilisation ‣ “Project overload” Easier to work on other Mental congestion when projects if knowledge is too much multitasking transferrable (Zika-Viktorsson, Sundstrom, Engwall, 2006) (Lindbeck and Snower, 2000)

Hardly any empirical evidence Rule of thumb (Weinberg, 1992) - not based on data 100 Working time available per project From: http://blog.codinghorror.com/the-multi-tasking-myth/ Loss to context switching 80 Percent of time 60 40 20 0 1 2 3 4 5 Number of simultaneous projects

Hardly any empirical evidence Rule of thumb (Weinberg, 1992) - not based on data 100 Working time available per project From: http://blog.codinghorror.com/the-multi-tasking-myth/ Loss to context switching 80 Percent of time 60 40 20 0 1 2 3 4 5 Number of simultaneous projects

Hardly any empirical evidence Rule of thumb (Weinberg, 1992) - not based on data 100 Working time available per project From: http://blog.codinghorror.com/the-multi-tasking-myth/ Loss to context switching 80 Percent of time 60 40 20 0 1 2 3 4 5 Number of simultaneous projects Recent work: ‣ Resuming interrupted tasks ‣ Work fragmentation (Parnin and DeLine, 2010) (Sanchez, Robbes, and Gonzalez, 2015)

Hardly any empirical evidence … but lots of data to test theories on. 14 million 35 million people projects

This work: Large-scale empirical study WHAT? Multitasking across projects ? ? ? ? Trends Reasons Effects Limits HOW? Sample: ‣ 1,200 programmers ‣ 5+ years of activity + Data mining User survey ‣ 50,000+ projects total (15% resp. rate)

This work: Large-scale empirical study Software developers multitask too WHAT? Multitasking across projects Trends & Reasons: EXAMPLE: GitHub developer (25 Nov 2013 — 18 May 2014) Details in paper #Projects 0 1 3 5 8 ? ? ? ? Trends Reasons Effects Limits Mon Tue Wed Thu Fri HOW? Sat Sun Nov Dec Jan Feb Mar Apr Sample: WHY? ‣ 1,200 programmers ‣ Request from other ‣ Being “stuck” ‣ Personal interest dev’s / management ‣ 5+ years of activity + Data mining User survey ‣ Signaling ‣ Downtime ‣ Dependencies ‣ 50,000+ projects total (15% resp. rate) Working for free? Motivations of participating The open source software development phenomenon: Activity traces and signals in software • • • in open source projects An analysis based on social network theory developer recruitment and hiring A. Hars and S. Ou. HICSS 2001 G. Madey, V. Freeh, and R. Tynan. AMCIS 2002 J Marlow, L Dabbish. CSCW 2013

Effects: perception vs. data “When contributing to multiple projects in parallel, I:” PERCEPTION Response Strongly disagree Disagree ree Neutral utral Agree Strongly agree increase project success ccess 15% 37% 47% resolve more issues ssues 23% 36% 40% feel more productive ctive 29% 37% 33% contribute more code overall verall 31% 40% 29% review more pull requests ests 34% 43% 23% introduce fewer bugs ugs 52% 43% 5% 100 50 0 50 100 Percentage

Effects: perception vs. data “When contributing to multiple projects in parallel, I:” PERCEPTION Response Strongly disagree Disagree ree Neutral utral Agree Strongly agree ccess 15% 37% 47% ssues 23% 36% 40% feel more productive ctive 29% 37% 33% verall 31% 40% 29% ests 34% 43% 23% ugs 52% 43% 5% 100 50 0 50 100 Percentage

Effects: perception vs. data “When contributing to multiple projects in parallel, I:” PERCEPTION Response Strongly disagree Disagree ree Neutral utral Agree Strongly agree ccess 15% 37% 47% resolve more issues ssues 23% 36% 40% ctive 29% 37% 33% contribute more code overall verall 31% 40% 29% review more pull requests ests 34% 43% 23% ugs 52% 43% 5% 100 50 0 50 100 Percentage

Effects: perception vs. data “When contributing to multiple projects in parallel, I:” PERCEPTION Response Strongly disagree Disagree ree Neutral utral Agree Strongly agree increase project success ccess 15% 37% 47% ssues 23% 36% 40% ctive 29% 37% 33% verall 31% 40% 29% ests 34% 43% 23% introduce fewer bugs ugs 52% 43% 5% 100 50 0 50 100 Percentage