AIH 2012




                                   If	
  you	
  fire	
  together,	
  you	
  wire	
  together	
  
                                        Prajni	
  Sadananda1,	
  Ramakoti	
  Sadananda2,	
  3	
  
       1
       	
     Department	
  of	
  Anatomy	
  and	
  Neuroscience,	
  University	
  of	
  Melbourne	
  Australian,	
  Australia	
  
                                      prajni.sadananda@unimelb.edu.au
                     2	
  
                         Institute	
  for	
  Integrated	
  and	
  Intelligent	
  Systems,	
  Griffith	
  University,	
  Australia	
  
                                                            3
                                                             NICTA,	
  Sydney	
  Australia	
  
                                            rsadananda@griffith.edu.au
	
  
       The	
  intention	
  of	
  this	
  paper	
  is	
  to	
  stimulate	
  discussion	
  on	
  Hebb’s	
  Law	
  and	
  
its	
  pedagogic	
  implications.	
  
       	
  
       At	
   a	
   basic	
   cellular	
   level,	
   Hebb’s	
   Law	
   states	
   that	
   is	
   Cell	
   A	
   and	
   Cell	
   B	
   persis-­‐
tently	
   fire,	
   the	
   connection	
   between	
   them	
   strengthens.	
   Figure	
   1	
   illustrates	
  
the	
   interactions.	
   This	
   is	
   a	
   cellular	
   levels	
   process,	
   suggesting	
   that	
   brain	
   pro-­‐
cesses	
  that	
  occur	
  repeartedly	
  tend	
  to	
  become	
  grafted	
  together	
  [1].	
  	
  
       	
  




                                                                                                                                                      	
  
Fig	
  1.	
  Hebb’s	
  Law.	
  Repeated	
  stimulation	
  results	
  in	
  a	
  stronger	
  signal	
  	
  




                                                                                                                                                              89
AIH 2012




           	
  

           	
  

                  This	
  scientific	
  theory	
  explains	
  the	
  adaptation	
  of	
  neurons	
  during	
  the	
  learn-­‐
           ing	
   process.	
   Importantly,	
   this	
   type	
   of	
   plasticity	
   does	
   not	
   involve	
   increasing	
  
           the	
  number	
  of	
  cells,	
  but	
  rather	
  strengthening	
  the	
  existing	
  cells’	
  connectivity.	
  
           Understanding	
   such	
   biological	
   phenomena	
   opens	
   up	
   new	
   paradigms	
   and	
  
           laws	
  that	
  AI	
  can	
  utilise.	
  The	
  question	
  is	
  whether	
  Hebb’s	
  law	
  would	
  stand	
  at	
  a	
  
           higher	
  level	
  of	
  abstraction?	
  There	
  are	
  suggestive,	
  but	
  not	
  conclusive	
  indica-­‐
           tions.	
  For	
  example,	
  a	
  friendship	
  is	
  considered	
  stronger	
  with	
  time,	
  indicating	
  
           a	
  strengthening	
  of	
  wiring	
  between	
  the	
  friends.	
  
                  	
  
                  Models	
   based	
   on	
   “firing	
   together	
   to	
   wire	
   together”	
   have	
   been	
   suggested	
  
           in	
  health	
  and	
  therapy	
  [2].	
  For	
  example,	
  if	
  a	
  patient	
  presents	
  with	
  a	
  mental	
  
           trauma	
   that	
   causes	
   extreme	
   anger,	
   the	
   therapist	
   introduces	
   a	
   counter	
   and	
  
           positive	
   stimulus	
   that	
   occurs	
   whenever	
   the	
   anger	
   occurs.	
   Both	
   (anger	
   and	
  
           the	
   positive	
   stimulus)	
   are	
   repeated	
   over	
   and	
   over	
   again,	
   thus	
   following	
  
           Hebb’s	
  Law	
  and	
  adding	
  strength	
  to	
  this	
  connection	
  between	
  the	
  two	
  stimuli,	
  
           resulting	
  in	
  relief	
  to	
  the	
  patient.	
  	
  
                  	
  
                  This	
  also	
  implies	
  causal	
  and	
  temporal	
  conjectures	
  based	
  on	
  causality.	
  The	
  
           causality	
  is	
  in	
  the	
  firing	
  sequence;	
  that	
  if	
  A	
  fires	
  first	
  and	
  then	
  B	
  fires,	
  A	
  is	
  the	
  
           cause.	
   If	
   B	
   fires	
   before	
   A,	
   a	
   reverse	
   interpretation	
   is	
   possible	
   that	
   may	
   de-­‐
           crease	
   the	
   strength	
   between	
   them.	
   There	
   seems	
   evidence	
   to	
   suggest	
   that	
  
           the	
  “firing”	
  and	
  “wiring”	
  may	
  be	
  a	
  sequential	
  process.	
  	
  
                  	
  
                  Causality	
  is	
  a	
  subject	
  of	
  intense	
  philosophical	
  interest	
  from	
  ancient	
  times.	
  
           Most	
   causal	
   models	
   are	
   rule-­‐based	
   systems.	
   They	
   demand	
   descriptions	
   of	
  
           the	
  world	
  at	
  two	
  points	
  in	
  time	
   –	
  a	
  before	
  and	
  an	
  after.	
  Two	
  problems	
  arise	
  
           here:	
   the	
   practical	
   computational	
   compulsions	
   make	
   these	
   rules	
   crudely	
  
           simplistic.	
  In	
  addition,	
  it	
  is	
  challenging	
  to	
  incorporate	
  temporal	
  effects	
  within	
  
           the	
  framework	
  of	
  rule	
  based	
  systems.	
  Hebb’s	
  law,	
  while	
  suggesting	
  causali-­‐
           ty,	
  does	
  not	
  provide	
  any	
  quantification.	
  Thus,	
  it	
  is	
  unlikely	
  that	
  an	
  alternative	
  
           formulation	
   of	
   causation	
   would	
   emerge	
   from	
   Hebb’s	
   law	
   alone.	
   We	
   may	
  
           look	
  for	
  another,	
  additional	
  neural	
  network	
  perspective	
  of	
  causation	
  here.	
  
                  	
  
                  Nevertheless,	
  causality	
  as	
  implied	
  with	
  Hebb’s	
  law	
  has	
  been	
  used	
  in	
  sci-­‐
           entific	
  research	
  and	
  therapeutics	
  to	
  a	
  large’	
  extent.	
  For	
  example,	
  oftentimes	
  
           doctors	
  complain	
  of	
  their	
  patients	
  being	
  unable	
  to	
  add	
  minor	
  and	
  incremen-­‐
           tal	
  changes	
  in	
  their	
  daily	
  routines	
  (such	
  as	
  exercise).	
  Understanding	
  Hebb’s	
  
           law	
  will	
  open	
  new	
  insights	
  into	
  why	
  this	
  might	
  be	
  so.	
  It	
  is	
  possible	
  that	
  the	
  
           patient	
   is	
   not	
   yet	
   “wired”	
   in	
   this	
   activity	
   and	
   requires	
   more	
   “firing”	
   before	
  




90
                                                                                                                                                 AIH 2012




                                                                                                                                          	
  

	
  

these	
  changes	
  can	
  be	
  established.	
  An	
  avenue	
  for	
  AI	
  research	
  is	
  to	
  aide	
  in	
  the	
  
development	
  of	
  tools	
  to	
  help	
  such	
  people	
  to	
  “re-­‐wire”.	
  	
  
	
  
       Indeed,	
   such	
   tools	
   exist	
   to	
   some	
   extent	
   to	
   treat	
   spinal	
   cord	
   injured	
   pa-­‐
tients	
   who	
   have	
   lost	
   motor	
   control	
   of	
   their	
   limbs.	
   In	
   a	
   non-­‐injured	
   situation,	
  
the	
  brain	
  delivers	
  pulses	
  to	
  the	
  lower	
  limbs	
  in	
  a	
  rhythmic/patterned	
  fashion	
  
to	
   allow	
   walking	
   action.	
   Once	
   a	
   spinal	
   injury	
   occurs,	
   the	
   connectivity	
   from	
  
the	
   brain	
   to	
   the	
   limbs	
   is	
   lost,	
   thereby	
   leaving	
   the	
   patient	
   immobile.	
   Stimula-­‐
tors	
  are	
  often	
  placed	
  below	
  the	
  level	
  of	
  the	
  injury,	
  which	
  deliver	
  patterned	
  
pulses	
   in	
   a	
   similar	
   manner	
   to	
   what	
   the	
   brain	
   was	
   previously	
   doing.	
   Over	
   a	
  
period	
  of	
  time,	
  a	
  spinal	
  pattern	
  generator	
  emerges,	
  which	
  thus	
  allows	
  some	
  
motion	
   of	
   the	
   lower	
   limbs	
   [3].	
   This	
   area	
   of	
   research	
   is	
   as	
   yet	
   in	
   its	
   infancy	
  
and	
  calls	
  for	
  a	
  better,	
  more	
  intelligent	
  systems	
  to	
  aide	
  these	
  patients.	
  
	
  
Conclusions:	
  	
  
       Artificial	
   intelligence	
   in	
   health	
   opens	
   up	
   chapters	
   of	
   great	
   opportunities	
  
and	
   exciting	
   challenges.	
   The	
   logical	
   calculus	
   articulated	
   by	
   McCulloch	
   and	
  
Pitts	
   [4]	
   forms	
   the	
   initial	
   basis	
   for	
   both	
   Symbolic	
   and	
   Connectionist	
   AI.	
   Since	
  
then	
  a	
  number	
  of	
  paradigms	
  have	
  emerged	
  on	
  all	
  aspects	
  of	
  AI	
  and	
  relating	
  
to	
  health	
  and	
  health	
  care.	
  The	
  emergence	
  of	
  the	
  convergence	
  of	
  computing	
  
and	
   communication	
   provides	
   us	
   boundless	
   opportunities	
   to	
   exploit	
   these	
  
paradigms	
  and	
  discover	
  the	
  new	
  ones.	
  
	
  
	
  

ReferenceƐ:	
  	
  

1.      Hebb,	
   D.	
   O.:	
   Organization	
   of	
   Behavior:	
   a	
   Neuropsychological	
   Theory.	
  
        John	
  Wiley,	
  New	
  York	
  (1949).	
  
2.      Atkinson,	
  B.,	
  Atkinson,	
  L.,	
  Kutz,	
  P.,	
  Lata,	
  L.,	
  Lata,	
  K.W.,	
  Szekely,	
  J.,	
  Weiss,	
  
        P.:	
  Rewiring	
  Neural	
  States	
  in	
  Couples	
  Therapy:	
  Advances	
  from	
  Affective	
  
        Neuroscience.	
  In:	
  Journal	
  of	
  Systemic	
  Therapies.	
  24,	
  3-­‐13	
  (2005)	
  
3.      Edgerton,	
   V.R.,	
   Roy,	
   R.R.:	
   A	
   new	
   age	
   for	
   rehabilitation.	
   Eur	
   J	
   Phys	
   Re-­‐
        habil	
  Med.	
  48,	
  99-­‐109	
  (2012)	
  
4.      McCulloch,	
  W.S.,	
  Pitts,	
  W.:	
  A	
  logical	
  Calculus	
  of	
  the	
  ideas	
  immanent	
  in	
  
        nervous	
   activity,	
   Bulletin	
   of	
   	
   mathematical	
   	
   Biophysics.	
   5,	
   115-­‐137	
  
        (1943).	
  	
  	
  

	
  




                                                                                                                                                      91