“On the Personality Traits of StackOverflow Users” - 29th IEEE ...

bolivialodgeInternet και Εφαρμογές Web

14 Δεκ 2013 (πριν από 3 χρόνια και 3 μήνες)

228 εμφανίσεις

29th  IEEE  Interna,onal  Conference  on  So3ware  Maintenance    
22  -­‐  28  September  2013  -­‐  Eindhoven,  The  Netherlands    

On  the  Personality  Traits  of    
 StackOverflow  Users

 
 
Blerina  Bazelli
,  Abram  Hindle,  Eleni  Stroulia  
 
Department  of  Compu,ng  Science  
 
University  of  Alberta,  Canada  
29th  IEEE  Interna,onal  Conference  on  So3ware  Maintenance    
22  -­‐  28  September  2013  -­‐  Eindhoven,  The  Netherlands    


So3ware  is  becoming  increasingly  “social”    


Developers  interact  through  social  networks  of  
experts  


How  do  personality  traits  impact  this  interac,on?    


Also  examined  by  Rigby  et  al.  
Mo,va,on  
2  
29th  IEEE  Interna,onal  Conference  on  So3ware  Maintenance    
22  -­‐  28  September  2013  -­‐  Eindhoven,  The  Netherlands    
Research  Ques,ons  
RQ1:  
 
Do  the  
top  reputed  authors
’  personality  types  differ?  
 
R
Q2:  
 
Do  author  personali,es  vary  by  
reputa1on
?  
 
RQ3
:
 
Do  authors  of  posts  belonging  to  the  same  
topics
 
share  similar  personality  types?  
 
R
Q4:
 
Do  personality  traits  have  an  impact  on  the  process  
of  
up-­‐vo1ng/down-­‐vo1ng  
posts?  
 
3  
29th  IEEE  Interna,onal  Conference  on  So3ware  Maintenance    
22  -­‐  28  September  2013  -­‐  Eindhoven,  The  Netherlands    
Aug.  2008  –  Aug.  2012  
~10  million  StackOverflow  posts
 
The  Data  Set
 
4  
29th  IEEE  Interna,onal  Conference  on  So3ware  Maintenance    
22  -­‐  28  September  2013  -­‐  Eindhoven,  The  Netherlands    
“Big  Five  Personality  Traits”
 
5  
29th  IEEE  Interna,onal  Conference  on  So3ware  Maintenance    
22  -­‐  28  September  2013  -­‐  Eindhoven,  The  Netherlands    
5  
LIWC  Dimension  
Your  Data  
Personal  Texts  
Formal  Texts  
Self-­‐references  
7.50  
11.4  
4.2  
Social  words  
10.00  
9.5  
8.0  
Posi,ve  Emo,ons  
5.00  
2.7  
2.7  
Nega,ve  emo,ons  
0.00  
2.6  
1.6  
Cogni,ve  words  
22.50  
7.8  
5.4  
Ar,cles  
2.50  
6.0  
7.2  
Big  words  
27.50  
13.1  
19.6  
…  
…  
…  
…  
Dic,onary  of  2700  English  words  
70  Dis,nct  Dimensions  (e.g.,  posi,ve  emo,ons)      
I,  me,  mine  
talk,  together  
love,  sweet  
hate,  dislike  
think,  believe,  
could  
Linguis,c  Inquiry  and  Word  Count  (LIWC)    
 

I  tried  to  understand  the  u,lity  of  backbone.js  from  its  site,  but  I  s,ll  
couldn't  figure  out  much.  Can  anybody  help  me  by  explaining  how  it  
works  and  how  could  it  be  helpful  in  wri,ng  beker  JavaScript?

 
6  
Compu,ng  the  Personality  Traits    
7  
Neuro1cism
 
 
=  First  Person  Singular  –  Ar,cles  –  Posi,ve  Emo,ons  +  Nega,ve  Emo,ons  
Agreeableness
 
 
 
=  First  Person  Singular  –  Big  Words  +  Posi,ve  Emo,on  –  Nega,ve  Emo,ons  
Openness
 
 
   
=  …  
…  
Qs  :
 
I  
tried
 to  understand  the  u,lity  of  backbone.js  from  its  site,  but  I  s,ll  
couldn't
 figure  out  much.  Can  
 
anybody  help  me  by  explaining  
how
 it  works  and  
how
 
could
 
it  be  helpful  in  wri,ng  beker  JavaScript?  
Ans:
 
Backbone.js  is  basically  an  uber-­‐light  framework  that  allows  you  to  structure  your  Javascript  code  in  an  
 
MVC  (Model,  View,  Controller)  fashion  where…  Some  
pros
 
that  I  discovered  with  Backbone:  extremely  
 
useful
 
underscore  u,lity  library,  code  is  
well
 
documented  and  a  
great
 
read…  There  is  also  this  
 
wonderful
 
Collec,on  class  that  lets  you  deal  with  collec,ons  of  models…  
 
 
 
 
Neuro1cism  
Extroversion  
Openness  
Agreeableness  
Conscien1ousness  
Ques,on  
5.41  
-­‐2.7  
15.32  
-­‐1.81  
-­‐3.60  
Answer  
0.00  
8.12  
21.60  
10.82  
-­‐13.51  
Pennebaker  et  al.  
29th  IEEE  Interna,onal  Conference  on  So3ware  Maintenance    
22  -­‐  28  September  2013  -­‐  Eindhoven,  The  Netherlands    
Preprocessing  
-­‐Discard  code  snippets  
-­‐Remove  HTML  tags  
Stack  
Overflow  
Posts    
LIWC  
Personality  traits’  
values  
H
0
:  
The  means  are  equal  
H
1
:  
The  means  are  not  equal  
ANOVA  
H
0  
H
1  
Tukey’s
 HSD  
The  authors  share  
similar  personality  traits  
Authors  who  share  
similar  personality  traits  
Methodology
 
8  
RQ1:  Do  the  
top  reputed  authors
’  personality  types  differ?  
9  
 95%  Confidence  Intervals  of  difference  of  means  
29th  IEEE  Interna,onal  Conference  on  So3ware  Maintenance    
22  -­‐  28  September  2013  -­‐  Eindhoven,  The  Netherlands    
10  
 95%  Confidence  Intervals  of  difference  of  means  
Top  vs.  
Medium  vs.  
Medium  
Low  
Low  
R
Q2:
 
Do  author  personali,es  vary  by  
reputa1on
?  
29th  IEEE  Interna,onal  Conference  on  So3ware  Maintenance    
22  -­‐  28  September  2013  -­‐  Eindhoven,  The  Netherlands    
0  
+  
-­‐  
11  
R
Q2:
 
Do  author  personali,es  vary  by  
reputa1on
?  
Top  vs.  Medium  
29th  IEEE  Interna,onal  Conference  on  So3ware  Maintenance    
22  -­‐  28  September  2013  -­‐  Eindhoven,  The  Netherlands    
 95%  Confidence  Intervals  of  difference  of  means  
Top  vs.  
Medium  vs.  
Medium  
Low  
Low  
R
Q2:  
 
Do  author  personali,es  vary  by  
reputa1on
?
 
12  
C#  vs.  
Android  vs.  
Java  vs.  
PHP  
 95%  Confidence  Intervals  of  difference  of  means  
Android  
Java  
JavaScript  
PHP  
Java  
JavaScript  
PHP  
JavaScript  
PHP  
JavaScript  vs.  
RQ3
:
 
Do  authors  of  posts  belonging  to  the  same  
topics
 
share  similar  personality  types?
 
13  
29th  IEEE  Interna,onal  Conference  on  So3ware  Maintenance    
22  -­‐  28  September  2013  -­‐  Eindhoven,  The  Netherlands    
Down-­‐voted  vs.    
Up-­‐voted  
ques1ons  
Down-­‐voted  vs.    
Up-­‐voted  
answers  
R
Q4:
 
Do  personality  traits  have  an  impact  on  the  
process  of  
up-­‐vo1ng/down-­‐vo1ng  
posts?
 
14  
29th  IEEE  Interna,onal  Conference  on  So3ware  Maintenance    
22  -­‐  28  September  2013  -­‐  Eindhoven,  The  Netherlands    
Conclusions
 


We  used  LIWC  to  explore  the  personality  traits  of  StackOverflow  
authors  


top  
reputed  
authors  are  less  
neuro,c,  
more  extroverted  and  
open    


authors  of  up-­‐voted  ques,ons  are  less  neuro,c  


Maybe  managers  could  use  these  insights  to  hire  more  

effec,vely  social

 programmers  
Thank  you!