ibm cog comp_may2015_handanimate

52
Is Your Intelligent Computer Smart Enough To Be on My Team? Barbara J. Grosz School of Engineering and Applied Sciences Harvard University Acknowledgements: Nuance FoundaEon

Upload: nancy-amoroso

Post on 09-Aug-2015

194 views

Category:

Documents


0 download

TRANSCRIPT

Is  Your  Intelligent  Computer    Smart  Enough  

To  Be  on  My  Team?  

Barbara  J.  Grosz  School  of  Engineering  and  Applied  Sciences  

Harvard  University    

Acknowledgements:  Nuance  FoundaEon        

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Background:  A  Par@cular    Turing  Predic@on  

}  “.  .  .  The  original  quesEon,  ‘Can  machines  think?’  I  believe  to  be  too  meaningless  to  deserve  discussion.  Nevertheless  I  believe  that  at  the  end  of  the  century  the  use  of  words  and  general  educated  opinion  will  have  altered  so  much  that  one  will  be  able  to  speak  of  machines  thinking  without  expec5ng  to  be  contradicted.  .  .”  

   

“Computing Machinery and Intelligence”, Mind, 1950. p. 442

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

What  Ques@on  Might  Turing  Pose  Now?    

}  SeOng:  Agents  and  people  work  together  on  nontrivial  task,    extended  in  Eme,  in  an  uncertain,  dynamic  environment.    

 }  Can  a  computer  (agent)  team-­‐member  behave,  over  the  long-­‐term,  in  such  a  way  that  people  on  the  team  will  not  noEce  it’s  not  human  or  think  it’s  stupid.    

Grosz,  Turing  Research  Symposium,  University  of  Edinburgh  &  Royal  Society  of  Edinburgh,  11  May  2012;  AI  Magazine,  33:4,  2012;  The  Atlan2c  blog,  August  2012.  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Why  Teamwork?  Canonical  Compu@ng  Environment  Now  

From “Computing Machinery and Intelligence” and “Can Machines Think?” to Teamwork

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Why  Teamwork?  Arguments  from  Cogni@ve  Science    }  It’s  fundamental:  “Every  funcEon  in  the  child's  cultural  development  appears  twice:  first,  on  the  social  level,  and  later,  on  the  individual  level  .  .  .    All  the  higher  func5ons  originate  as  actual  rela5ons  between  human  individuals.”  (Vygotsky,  1978,  pp.  56-­‐57)  

 }  Language  is  interpersonal,  cooperaEve:  

}  “Infants  depend  in  both  their  comprehension  and  producEon  of  poinEng  on  a  joint  ahenEonal  frame  (common  ground)  with  their  communicaEve  partners  .  .  .  And  they  do  this  for  the  fundamentally  coopera5ve  mo5ves  of  helping  and  sharing  informaEon  and  aOtudes  .  .  .”  (Tomasello,  et  al.,  2007,  p.  720)  

 

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Teamwork  Is  Not  Easy:  Essen@al  Match  of  Abili@es  to  Tasks  and  Roles  or  .  .  .    

Source: http://msdn.microsoft.com

Interac@ve  Intelligence  and    Health  Care  Coordina@on  

new  twists:  loosely  coupled  team  members  

team  of  peers  (limited/no  hierarchy)  relaEvely  long  Eme  horizon  (not  “one-­‐shot”)  

 

My  current  research  focus:    

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Research  Program  Framing:  Design/Build    Collabora@ve  Computer  Agents  

Models  and  algorithms  for  efficient  decision-­‐making  under  uncertainty    Formal  models  of  collaboraEon  

Methods  for  supporEng  communicaEon  and  informaEon  sharing  

Methods  for  analyzing  and  learning  people’s  percepEons  of  agent  behavior  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Theore@cal  Founda@ons:  SharedPlans  Theory  }  Each  team  members  commits  to  team’s  performance  of  the  group  acEvity;  requires  model  of  inten2on.  

}  Team  members  reach  consensus  on  a  (high-­‐level)  “recipe”,  recipes  may  be  par2al,  revised  over  2me.    

}  Team  reaches  consensus  on  allocaEon  of  (subtasks),  taking  into  account  agents’  capabiliEes.    

}  Team  members  commit  to  assigned  subtasks.  }  Team  members  commit  to  each  others’  success.    

Theory  implicitly  requires  group  decision  making  processes.  

(Grosz&Kraus, 1996,1999; Hunsberger, 1999; Grosz&Hunsberger, 2006 )  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Theory  As  Inspira@on  and  Design  Guide:  Collabora@ve  Interfaces  I:  

What’s  the  right  division  of  labor?    

Writer’s  Aid  (Babaian, Grosz, Shieber, 2002)  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Theory  As  Inspira@on  and  Design  Guide:  Collabora@ve  Interfaces  II:  

Detailed  descripEon  of  students’  interacEons  

Condensed  presentaEon  of  students’  interacEons  

Direct  feedback  from  teacher  to  students  

Complemen2ng  and  working  well  with  people.   S-­‐CASTS  (Reddy et al., 2009; Gal, Reddy,

Shieber, Rubin, Grosz, 2012)  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Key  Problem:    Interrup@on  Management  

drive    home    

Route  A   Route  B  

0.9   0.1  

drive    home    

Route  A  

Agent  (Observe/Help)    

Recipient  Communicate?    

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Key  Problem:    Interrup@on  Management  

drive    home    

Route  A   Route  B  

0.9   0.1  

drive    home    

Route  A  

Inform  if  Expected  U@lity>Cost  Agent  

(Observe/Help)    

Recipient  Surprise!  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

A  New  Representa@on:    Probabilis@c  Recipe  Trees  (Kamar,  2010)  

May 20, 2015 Slide 14

Expected Duration:

19 min

min 10 min 5 min 3 min 10 min 14 min 2 min 2

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Inform  Decisions  

May 20, 2015 Ece Kamar - Harvard University

Slide 15

Expected Duration: 47.8 min

min 5 min 3 min 10 min 14 min 2 min 242 min

Route A

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Inform  Decisions  

May 20, 2015 Ece Kamar - Harvard University

Slide 16

Expected Duration:

28 min

min 10 min 14 min 2 min 2

0.11.0 0.0 0.0

Route A

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Interrupt  to  Inform?  EU( )

Updated Plan

EU( )

Original Plan

>  

Communicate  if  the  gain  is  bigger  than  the  cost  of  communicaEon  and  interrupEon  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

The  Vision  and  the  Challenges  }  “Turing  Challenge”:  Can  a  computer  (agent)  team-­‐member  behave,  over  the  long-­‐term,  in  such  a  way  that  people  on  the  team  will  not  no2ce  it’s  not  human  or  think  it’s  stupid.    

}  Test  Environment:  Health  Care  Coordina2on    }  Vision:  Care  Augmen2ng  SoQware  Partners  (CASPERs)  

}  AI  scienEfic  challenges  (MAS,  AI)  arise  from    }  demands  of  coordinaEng  care  }  needs  for  improving  communicaEon  between  paEents  and  medical  providers    

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Parents

Primary Care

Provider

Physical Therapist

Neurologist

School nurse

Camp counselor

Health aide

Teacher

Speech Therapist

GI

The  Care  for  Children  with  Complex  Chronic  Condi@ons  (Amir,  Grosz,  Law  &  Stern,  AAMAS  2013)  

Project  in  collabora2on  with  the  Complex  Primary  Care  Clinic,    Stanford  University  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Parents

Primary Care

Provider

Physical Therapist

Neurologist

School nurse

Camp counselor

Health aide

Teacher

Speech Therapist

GI

The  Care  for  Children  with  Complex  Chronic  Condi@ons  (Amir,  Grosz,  Law  &  Stern,  AAMAS  2013)  

The Problem: care for children with complex conditions is poorly coordinated,

leading to unmet health needs and preventable health care crises

Project  in  collabora2on  with  the  Complex  Primary  Care  Clinic,    Stanford  University  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Team-­‐Based  Care  Plans  for  Improved  Coordina@on  (LPFCH,  2014)  

Goals Actions Caregivers Move to oral feeds

•  Improve mouth muscle tone •  Adjust formula for weight

gain

PCP, GI, OT, nutritionist

Start daycare

•  Minimize need for tube feeds

•  Assess therapy needs

Parents, PCP, nutritionist, home nurse

Go on family trip

•  Arrange portable equipment •  Arrange funding and

transportation

Parents, PCP, PT, social worker

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Team-­‐Based  Care  Plans  for  Improved  Coordina@on  (LPFCH,  2014)  

Goals Actions Caregivers Move to oral feeds

•  Improve mouth muscle tone •  Adjust formula for weight

gain

PCP, GI, OT, nutritionist

Start daycare

•  Minimize need for tube feeds

•  Assess therapy needs

Parents, PCP, nutritionist, home nurse

Go on family trip

•  Arrange portable equipment •  Arrange funding and

transportation

Parents, PCP, PT, social worker

Rationale: everybody “on the same page” In practice: rarely deployed or consulted

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Study  of  Complex  Care  Teams  (Amir,  Grosz,  Gajos,  Swenson,  and  Sanders,  2015)  

}  “From  Care  Plans  to  Care  Coordina@on:  Opportuni@es  for  Computer  Support  of  Teamwork  in  Complex  Healthcare”    

}  Interviews  and  observaEons  of  team  members:  }  Parents  (13)  }  Primary  care  providers  (4)  }  Specialists  (4)  }  Therapists  (8)  }  Care  coordinator  (1)  }  Program  directors  (2)  }  Family  services  coordinator  (1)  }  Social  worker  (1)  

}  Analyzed  using  affinity  diagramming  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Barriers  to  Effec@ve  Care  Plan  Use:  Complex  Teamwork  in  Complex  Care  

“FLECS”  teamwork  characterisEcs:    }  Flat-­‐structure  of  team  }  Loosely  coupled  plans  and  acEviEes  }  Extended  duraEon  of  plans    }  ConEnual  distributed  revision  of  plans  }  Syncopated  Eme  scales  

 

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Flat  Structure  No  single  person  in  charge:  

“We  have  different  goals  for  different  specialists;  it  is  hard  to  keep  track.”  (parent)  

Need  to  priori2ze  goals  because  “everyone  wants  to  work  on  everything.”  (parent)    

 

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Loosely  Coupled  Ac@vi@es  

Loose  coupling  makes  appropriate  informa5on  sharing  hard:  

“There  isn’t  an  example  when  I  wasn’t  missing  informa2on”    (specialist)  

“We  need  to  relay  informa2on  back  and  forth...”  (parent)  

 

 

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Extended  Dura@on,  Con@nual  Distributed  Plan  Revision  

No  mechanism  to  support  plan  revision:    Full-­‐team  mee2ngs  “totally  not  scalable”  (specialist)    “All  the  status  chats  have  to  be  provider  ini2ated,  and  so  if  you  don’t  remember  to  do  it  or  there’s  no  one  coordina2ng  it,  it’s  like  where  is  it  going,  where  do  you  even  look  for  it?”  (specialist)    

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Syncopated  Time  Scales  

Different  frequencies  of  seeing  the  pa5ent  }  Primary  care  providers:    3  to  4  2mes  a  year  }  Specialists:  2  to  3  2me  a  year  }  Therapists:  1  to  3  2mes  a  week  

 Different  informa5on  needs:  “A  doctor  asks  if  she  is  walking  and  expects  a  yes/no  answer;  a  physical  therapist  will  ask  how  she  is  walking  and  how  much  progress  she  has  made.”  (parent)  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Team-­‐Based  Care  Plans:  Ideal  vs.  Reality  

}  Principles  for  successful  care  plan  use  (LPFCH,  2014):  }  “The  plan  of  care  is  systema2zed  as  a  common,  shared  document;  it  is  used  consistently  by  every  provider…”  

}  “The  team  monitors  progress  against  goals,  provides  feedback  and  adjusts  the  plan  of  care  on  an  ongoing  basis…”  

}  “Family-­‐centered  care  teams  can  access  the  informa5on  they  need  to  make  shared,  informed  decisions.”  

}  Principles  are  largely  not  realized  in  current  pracEce;  root  cause:  FLECS  nature  of  teamwork.    

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Team-­‐Based  Care  Plans:  Ideal  vs.  Reality  

}  Principles  for  successful  care  plan  use  (LPFCH,  2014):  }  “The  plan  of  care  is  systema2zed  as  a  common,  shared  document;  it  is  used  consistently  by  every  provider…”  

}  “The  team  monitors  progress  against  goals,  provides  feedback  and  adjusts  the  plan  of  care  on  an  ongoing  basis…”  

}  “Family-­‐centered  care  teams  can  access  the  informa5on  they  need  to  make  shared,  informed  decisions.”  

}  Principles  are  largely  not  realized  in  current  pracEce;  root  cause:  FLECS  nature  of  teamwork.    

Research focus: Develop foundations and technology to better support complex

FLECS teamwork.

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Founda@ons  for  Design  of  Systems  to  Support  Complex  Care  Teams  

SharedPlans  (Grosz  &  Kraus  1996)  :      A  computaEonal  theory  of  collaboraEon            “..the  capabili2es  needed  for  collabora2on  cannot  be  patched  on  but  must  be  designed  in  from  the  start.  "    

SharedPlans  Representa@on  

adjust  formula  

funding  &  transportaEon  

follow  family  prioriEes  

move  to  oral  feeds  

go  on  family  trip  

improve  mouth  muscle  tone  

{parents,  primary  care  provider,  specialists,  therapists,  community  members}  

{primary  care  provider,    physical  therapist,  social  worker}  

{primary  care  provider,  gastroenterologist,  occupa2onal  therapist,  nutri2onist}  

arrange  equipment  {…}   {…}   {…}   {…}  

SharedPlans  Representa@on  

adjust  formula  

funding  &  transportaEon  

follow  family  prioriEes  

move  to  oral  feeds  

go  on  family  trip  

improve  mouth  muscle  tone  

{parents,  primary  care  provider,  specialists,  therapists,  community  members}  

{primary  care  provider,    physical  therapist,  social  worker}  

{primary  care  provider,  gastroenterologist,  occupa2onal  therapist,  nutri2onist}  

arrange  equipment  {…}   {…}   {…}   {…}  

adjust  formula  

arrange  equipment  

go  on  family  trip  

move  to  oral  feeds  

SharedPlans Representation

funding  &  transportaEon  

follow  family  prioriEes  

{parents,  primary  care  provider,  specialists,  therapists,  community  members}  

{primary  care  provider,    physical  therapist,  social  worker}  

{primary  care  provider,  gastroenterologist,  occupa2onal  therapist,  nutri2onist}  

{…}   {…}   {…}   {…}  improve  mouth  muscle  tone  

improve  mouth  muscle  tone  

SharedPlans Representation

adjust  formula  

arrange  equipment  

go  on  family  trip  

move  to  oral  feeds  

funding  &  transportaEon  

follow  family  prioriEes  

{parents,  primary  care  provider,  specialists,  therapists,  community  members}  

{primary  care  provider,    physical  therapist,  social  worker}  

{…}   {…}   {…}   {…}  

{primary  care  provider,  gastroenterologist,  occupa2onal  therapist,  nutri2onist}  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Key  Elements  of  SharedPlans  Theory  }  Consensus  on  recipe:      

}  Recipes  may  be  parEal  and  evolve  over  Eme:  

}  Team  members  commit  to  performance  of  group  acEvity  and  to  each  others’  success:  

(Grosz&Kraus, 1996,1999; Hunsberger, 1999; Grosz&Hunsberger, 2006 )  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Key  Elements  of  SharedPlans  Theory  }  Consensus  on  recipe:      

}  Support  for  providers  establishing  agreement  on  high-­‐level  approach,  establishing  mutual  belief.  

}  Recipes  may  be  parEal  and  evolve  over  Eme:  }  Support  dynamically  evolving  plans.      

}  Team  members  commit  to  performance  of  group  acEvity  and  to  each  others’  success:  }  Support  communicaEon  and  coordinaEon  at  appropriate  levels  and  Emes.    

(Grosz&Kraus, 1996,1999; Hunsberger, 1999; Grosz&Hunsberger, 2006 )  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Key  Elements  of  SharedPlans  Theory  }  Consensus  on  recipe:      

}  Support  for  providers  establishing  agreement  on  high-­‐level  approach,  establishing  mutual  belief.  

}  Recipes  may  be  parEal  and  evolve  over  Eme:  }  Support  dynamically  evolving  plans.      

}  Team  members  commit  to  performance  of  group  acEvity  and  to  each  others’  success:  }  Support  communicaEon  and  coordinaEon  at  appropriate  levels  and  Emes.    Key  capability:  informaEon  sharing  without  informaEon  

overload.      

(Grosz&Kraus, 1996,1999; Hunsberger, 1999; Grosz&Hunsberger, 2006 )  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Key  Roles  for  Technology  for  Suppor@ng  Complex  Care  Teams  }  Make  the  care  plan  “ever  present”  

}  Support  plan  revision  and  expansion  

}  Support  efficient  informaEon  sharing  

Challenges:  

}  EliciEng  plans  

}  Inferring  context  in  plan  

}  Reasoning  about  informaEon  sharing  

 

adjust  formula  

Agreement on High-Level Approach, Mutual Beliefs

funding  &  transportaEon  

follow  family  prioriEes  

move  to  oral  feeds  

go  on  family  trip  

improve  mouth  muscle  tone  

{parents,  primary  care  provider,  specialists,  therapists,  community  members}  

{primary  care  provider,    physical  therapist,  social  worker}  

{primary  care  provider,  gastroenterologist,  occupa2onal  therapist,  nutri2onist}  

arrange  equipment  {…}   {…}   {…}   {…}  

Dynamically Evolving Plans

…   …   …   …  

funding  &  transportaEon  

follow  family  prioriEes  

move  to  oral  feeds  

go  on  family  trip  

improve  mouth  muscle  tone  

adjust  formula  

arrange  equipment  

{parents,  primary  care  provider,  specialists,  therapists,  community  members}  

{primary  care  provider,    physical  therapist,  social  worker}  

{primary  care  provider,  gastroenterologist,  occupa2onal  therapist,  nutri2onist}  

{…}   {…}   {…}   {…}  

adjust  formula  

…   …   …   …  

funding  &  transportaEon  

follow  family  prioriEes  

move  to  oral  feeds  

go  on  family  trip  

improve  mouth  muscle  tone  

{parents,  primary  care  provider,  specialists,  therapists,  community  members}  

{primary  care  provider,    physical  therapist,  social  worker}  

{primary  care  provider,  gastroenterologist,  occupa2onal  therapist,  nutri2onist}  

arrange  equipment  

Communication and Coordination

{…}   {…}   {…}   {…}  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Ongoing  Work:    GoalKeeper  (Ofra  Amir)  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Parents

PCP

Physical Therapist

GI

School nurse

Camp counselor

Health aide

Teacher

Speech Therapist

Neurologist

Beyond  Health  Care:  Plan  Coordina@on  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Victims

Police

Fire Fighter

National Guard

Politician

Business man

ICU

NGOs Construction

Coordina@ng  Rescue  and  Rebuilding  

Psychologist

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Ongoing  Work:  Suppor@ng  Collabora@ve  Wri@ng  

             

   “Deploying  AI  Methods  to  Support  Collabora2ve  Wri2ng:  a  Preliminary  

Inves2ga2on”,    Gehrmann,  Urke,  Amir  and  Grosz,  CHI  2015  

 

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

No  man  is  an  island,    en@re  of  itself...  (John  Donne,  1624)  

computer

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Giving  Turing  the  Last  Word:  Fundamental  Knowledge    }  “The  whole  thinking  process  is  sEll  rather  mysterious  to  us,  but  I  believe  the  ahempt  to  make  a  thinking  machine  will  help  us  greatly  in  finding  out  how  we  think  ourselves.”  

Typescript,  BBC  Third  Programme  15  May  1951;  reprinted  in  Shieber,  The  Turing  Test:  Verbal  Behavior  as  the  Hallmark  of  Intelligence,  2004  

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

References  SharedPlans:  Barbara  J.  Grosz  and  Luke  Hunsberger,  2006.  The  Dynamics  of  IntenEons  in  CollaboraEve  IntenEonality.      In  Cogni2ve  Systems  Research  (special  issue  on  Cogni2on,  Joint  Ac2on  and  Collec2ve  Inten2onality),  7,2-­‐3.      

Luke  Hunsberger.  1999.  “Making  SharedPlans  More  Concise  and  Easier  to  Reason  About.”  In  Intelligent  Agents  V:  Agents  Theories,  Architectures,  and  Languages,  pp.  81-­‐98.  Springer  Berlin  Heidelberg.    Barbara  Grosz  and  Sarit  Kraus.  1999.  The  EvoluEon  of  SharedPlans.  Founda2ons  of  Ra2onal  Agencies,  A.  Rao  and  M.  Wooldridge,  eds.,  Kluwer  Academic  Press,    pp.  227-­‐262.  

 Barbara  J.  Grosz  and  Sarit  Kraus.  1996.  CollaboraEve  Plans  for  Complex  Group  AcEon.  In  Ar2ficial  Intelligence  86(2),  pp.  269-­‐357.    Awarded  IFAAMAS  Influen2al  Paper  Award,  2007.      

SharedPlans  for  dialogue:  Karen  E.  Lochbaum.  1998.  A  collaboraEve  planning  model  of  intenEonal  structure.  Computa2onal  Linguis2cs.  24,  4  (December  1998),  525-­‐572.      

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

References  Collabora@ve  interfaces:    S-­‐CASTS:  Ya'akov  Gal,  Swapna  Reddy,  Stuart  Shieber,  Andee  Rubin,  and  Barbara  Grosz.  2012.  Plan  RecogniEon  in  Exploratory  Domains.  Ar2ficial  Intelligence.  176(1):  pp.  2270—2290.      WAID:  Tamara  Babaian,  Barbara  J.  Grosz  and  Stuart  M.  Shieber.    2002  A  Writer’s  CollaboraEve  Aid.  Proceedings  of  the  Intelligent  User  Interfaces  Conference  (IUI-­‐2002),  San  Francisco,  CA.    January  13-­‐16.    ACM  Press.    pp.  7-­‐14.      Colored  Trails  MAS  Testbed:    Ya’akov  Gal,  Barbara  Grosz,  Sarit  Kraus,  Avi  Pfeffer,  Stuart  Shieber.  2010.  Agent  Decision-­‐Making  in  Open  Mixed  Networks.  Ar2ficial  Intelligence,  174(18):  pp.  1460-­‐1480.      Interrup@on  Management:  Ece  Kamar,  Kobi  Gal,  and  Barbara  Grosz.  2013.  Modeling  InformaEon  Exchange  OpportuniEes  for  EffecEve  Human-­‐computer  Teamwork.  Ar2ficial  Intelligence  195:528–550.      Ece  Kamar,  Ya'akov  Gal,  and  Barbara  J.  Grosz.  2009.  IncorporaEng  helpful  behavior  into  collaboraEve  planning.  In  Proceedings  of  The  8th  Interna2onal  Conference  on  Autonomous  Agents  and  Mul2agent  Systems  -­‐  Volume  2  (AAMAS  '09),  Vol.  2.  InternaEonal  FoundaEon  for  Autonomous  Agents  and  MulEagent  Systems,  Richland,  SC,  pp.  875-­‐882.      

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

References  Informa@on  Sharing  for  Care  Coordina@on  and  Collabora@ve  Work:  Ofra  Amir,  Barbara  Grosz,  Krzysztof  Gajos,  Sonja  Swenson,  and  Lee  Sanders.  2015.  From  Care  Plans  to  Care  CoordinaEon:  OpportuniEes  for  Computer  Support  of  Teamwork  in  Complex  Healthcare.”  In  Proceedings  of  the  33rd  Annual  ACM  Conference  Extended  Abstracts  on  Human  Factors  in  Compu2ng  Systems.  Seoul,  South  Korea.  April  18-­‐23,  2015.  pp.  1419-­‐1428      Ofra  Amir,  Barbara  J.  Grosz,  Edith  Law,  and  Roni  Stern.  2013.  CollaboraEve  Health  Care  Plan  Support.  Proceedings  of  the  12th  Interna2onal  Conference  on  Autonomous  Agents  and  Mul2agent  Systems  (AAMAS  2013),  Ito,  Jonker,  Gini,  and  Shehory  (eds.).  pp.  793-­‐796.  Second  Prize,  Compu2ng  Community  Consor2um/AAMAS2013  Challenges  and  Visions  Track      SebasEan  Gehrmann,  Lauren  Urke,  Ofra  Amir,  and  Barbara  J  Grosz.  2015.  Deploying  AI  Methods  to  Support  CollaboraEve  WriEng:  A  Preliminary  InvesEgaEon.  In  Proceedings  of  the  33rd  Annual  ACM  Conference  Extended  Abstracts  on  Human  Factors  in  Compu2ng  Systems.  Seoul,  South  Korea.  April  18-­‐23,  2015.  pp.  917-­‐922.    

Barbara J. Grosz Smart Enough To Be On My Team

IBM CSIG May 2015

Thanks!    

Collaborators: Ofra Amir, Krzysztof Gajos Harvard SEAS; Sonja Swenson, Lee Sanders Stanford University Medical School.  Research supported in part by Nuance Foundation.

 QuesEons?