On  Using  Informa.on  Retrieval  for  the  Selec.on  &    Sensi.vity  Review  of  Digital  Public  Records  

Timothy  Gollins1,2  Graham  McDonald1  

Craig  Macdonald1  Iadh  Ounis1  

 1  University  of  Glasgow  UK,  2  The  Na.onal  Archives  UK  

Open  Government  

●  The  release  of  Public  Records  facilitate  accountability  under  the  rule  of  law  -  Includes  archival  of  records  and  freedom  of  informa0on  

●  To  be  effec.ve  they  must  be  reasonably  contemporaneous  and  open  

Transfer Permanent

Preservation Open

Presentation Selection

& Appraisal Sensitivity

Review

Government Department The National Archives

Public  Records  Transfer  Process  (UK)  

Transfer Permanent

Preservation Open

Presentation Selection

& Appraisal Sensitivity

Review

Government Department The National Archives

Public  Records  Transfer  Process  (UK)  

Selec=on  &  Appraisal    

●  Cannot  keep  everything  –  All  records  are  important,  but  some  more  than  others  

●  Archives  Catalogue  and  Selec.on  -  Based  on  context  of  crea0on  not  subject  

●  Breakdown  in  Admin  Prac.ces  when  Digital  records  introduced  –  Digital  Records  have  poor  contexts  

●  Need  tools  to  extract  and  confer  meaningful  structure  –  Based  on  the  Context  of  Crea0on  and  Distribu0on,  not  just  on  the  Content  

Sensi=vity  Review  ●  Sensi.ve  records  remain  closed  –  upto  120  years  ●  Possible  (challenging)  Sensi.vi.es    

–  Defence  and  Na*onal  Security  –  e.g.  Nuclear  missile  deployment  

–  Commercial  Confiden*ality  –  e.g.  Notes  of  contract  nego0a0ons  

–  Damage  to  Interna*onal  Rela*ons  –  e.g.  Insul0ng  remarks  about  a  leader  (past  or  even  the  current  one!)  

–  Personal  Privacy  and  Health  &  Safety  -­‐  e.g.  Religious  beliefs  or  names  of  informants    

●  Important  to  avoid  precau.onary  closure    –  Even  though  this  is  lawful,  it  is  Morally,  Ethically,  and  Poli0cally  unacceptable    

Sensi=vity  &  Privacy  

●  Personal  privacy  only  one  aspect  of  sensi.vity  ●  Sensi.vity  is  rather  like  personal  privacy,  but  of  countries  and  organisa.ons  

●  Common  Proper.es  (discovered  by  our  own  work)    –  Distributed  and  Diffused  within/across  documents  –  Concerned  with  context  (including  prior  knowledge)  –  May  not  depend  on  subject  maNer  of  record  

●  Study  of  sensi.vity  necessary  for  general  solu.ons  to  privacy  

Ways  Forward  &  Links  to  IR  (1)    

●  Deep  study  of  human  reviewers  –  To  understand  sensi0vity  itself  and  how  it  is  perceived  

•  Percep0on  of  Sensi0vity  vs.  Relevance  –  c.f.  [Webber  &  Pickens  SIGIR  2013]  

-  Order  of  presenta0on  for  review  –  c.f.  sensi0vity  is  o[en  concerned  with  narra0ve  –  c.f.  [Scholer  et  al  SIGIR  2013]  

–  For  feature  iden0fica0on  

●  Develop  new  features    –  For  context  –  c.f.  relevance  propaga0on  [Qin  et  al  SIGIR  2005]  –  and  diffuse  proper0es      

Ways  Forward  &  Links  to  IR  (2)  

●  Use  correla.on  with  external  informa.on  ●  Extension  of  test  collec.on  

–  Challenging  given  the  nature  of  sensi0vity  

•  Assistance  not  Automa.on  –  Reviewer  must  jus0fy  in  law,  therefore  require  assistance  –  c.f.  [Azzopardi  SIGIR  2011]    

●  Use  ac.ve  machine  learning  to  adjust  for  local  circumstances  

Conclusion  

●  Transi.on  to  Digital  Records  enables  beTer  access  in  principle  (c.f.  online),  but  presents  challenges  that  puts  open  government  at  risk  

●  Sensi.vity  offers  great  opportuni.es  and  challenges  for  both  IR  research  and  many  domains  beyond  (e.g.  Social  Science,  Poli.cs,  etc.)  

●  Sensi.vity  and  Privacy  issues  need  much  more  than  technical  solu.ons    

 ●  Study  of  sensi.vity  are  necessary  for  general  solu.ons  to  privacy  (also  for  selec.on  &  appraisal)  

     

10