UNDERSTANDING IS A POOR SUBSTITUTE FOR CONVEXITY (ANTIFRAGILITY) Something cent !"# $e % cent !"# i& mi&&ing in hi&to ic!

" !cco'nt& o( &cienti(ic !n) techno"ogic!" )i&co$e %* The )i&co' &e !n) cont o$e &ie& (oc'& on the o"e o( "'c+ !& o,,o&e) to te"eo"ogic!" , og !m& (( om telos, -!im-)# th!t i&# one& th!t e"% on , e.&et )i ection ( om (o m!" &cience* Thi& i& ! (!'/.)e0!te1 "'c+ c!nnot "e!) to (o m!" e&e! ch ,o"icie&2 one c!nnot &%&tem!ti3e# (o m!"i3e# !n) , og !m !n)omne&&* The ) i$e i& neithe "'c+ no )i ection# 0't m'&t 0e in the !&%mmet % (o con$e/it%) o( ,!%o((&# ! &im,"e m!them!tic!" , o,e t% th!t h!& "ie) hi))en ( om the )i&co' &e# !n) the 'n)e &t!n)ing o( 4hich c!n "e!) to , eci&e e&e! ch , inci,"e& !n) , otoco"&* 5ISSING T6E ASY55ETRY The "'c+ $e &'& +no4"e)ge &to % i& !& (o""o4&* I onic!""%# 4e h!$e $!&t"% mo e e$i)ence (o e&'"t& "in+e) to "'c+ th!n to tho&e coming ( om the te"eo"ogic!"# o't&i)e ,h%&ic&7e$en !(te )i&co'nting (o the &en&!tion!"i&m* In &ome o,!8'e !n) non"ine! (ie")&# "i+e me)icine o enginee ing# the te"eo"ogic!" e/ce,tion& ! e in the mino it%# &'ch !& ! &m!"" n'm0e o( )e&igne ) 'g&* Thi& m!+e& '& "i$e in the cont !)iction th!t 4e "! ge"% got he e to 4he e 4e ! e th!n+& to 'n)i ecte) ch!nce# 0't 4e 0'i") e&e! ch , og !m& going (o 4! ) 0!&e) on )i ection !n) n! !ti$e&* An)# 4h!t i& 4o &e# 4e ! e ('""% con&cio'& o( the incon&i&tenc%* The ,oint 4e 4i"" 0e m!+ing he e i& th!t "ogic!""%# neithe t i!" !n) e o no -ch!nce- !n) &e en)i,it% c!n 0e 0ehin) the g!in& in techno"og% !n) em,i ic!" &cience !tt i0'te) to them* B% )e(inition ch!nce c!nnot "e!) to "ong te m g!in& (it 4o'") no "onge 0e ch!nce)2 t i!" !n) error c!nnot 0e 'ncon)ition!""% e((ecti$e1 e o & c!'&e ,"!ne& to c !&h# 0'i")ing& to co""!,&e# !n) +no4"e)ge to eg e&&* The 0ene(ici!" , o,e tie& h!$e to e&i)e in the t%,e o( e/,o&' e# th!t i&# the ,!%o(( ('nction !n) not in the -"'c+- ,! t1 the e nee)& to 0e ! &igni(ic!nt !&%mmet % 0et4een the g!in& (!& the% nee) to 0e "! ge) !n) the e o & (&m!"" o h! m"e&&)# !n) it i& ( om &'ch !&%mmet % th!t "'c+ !n) t i!" !n) e o c!n , o)'ce e&'"t&* The gene !" m!them!tic!" , o,e t% o( thi& !&%mmet % i& con$e/it% (4hich i& e/,"!ine) in Fig' e 9)2 ('nction& 4ith "! ge g!in& th!n "o&&e& ! e non"ine! .con$e/ !n) e&em0"e (in!nci!" o,tion&* C itic!""%# con$e/ ,!%o((& 0ene(it ( om 'nce t!int% !n) )i&o )e * The non"ine! , o,e tie& o( the ,!%o(( ('nction# th!t i&# con$e/it%# !""o4 '& to (o m'"!te !tion!" !n) igo o'& e&e! ch ,o"icie&# !n) one& th!t !""o4 the h! $e&ting o( !n)omne&&*

Figure 1- More Gain than Pain from a Random Event. The performance curves outward, hence loo s !conve"!. #n$where where such as$mmetr$ prevails, we can call it conve", otherwise we are in a concave position. The implication is that $ou are harmed much less %$ an error &or a variation' than $ou can %enefit from it, $ou would welcome uncertaint$ in the long run.

OPA:UE SYSTE5S AND OPTIONALITY F' the # it i& in com,"e/ &%&tem&# one& in 4hich 4e h!$e "itt"e $i&i0i"it% o( the ch!in& o( c!'&e. con&e8'ence&# th!t tin+e ing# 0 ico"!ge# o &imi"! $! i!tion& o( t i!" !n) e o h!$e 0een &ho4n to $!&t"% o't,e (o m the te"eo"ogic!"7it i& n!t' e;& mo)'& o,e !n)i* B't tin+e ing nee)& to 0e con$e/2 it i& im,e !ti$e* T!+e the mo&t o,!8'e o( !""# coo+ing# 4hich e"ie& enti e"% on the he' i&tic& o( t i!" !n) e o # !& it h!& not 0een ,o&&i0"e (o '& to )e&ign ! )i&h )i ect"% ( om chemic!" e8'!tion& o e$e &e. enginee ! t!&te ( om n't ition!" "!0e"&* <e t!+e h'mm'&# !)) !n ing e)ient# &!% ! &,ice# t!&te to &ee i( the e i& !n im, o$ement ( om the com,"e/ inte !ction# !n) et!in i( 4e "i+e the !))ition o )i&c! ) the e&t* C itic!""% 4e h!$e the o,tion# not the o0"ig!tion to +ee, the e&'"t# 4hich !""o4& '& to et!in the ',,e 0o'n) !n) 0e 'n!((ecte) 0% !)$e &e o'tcome&* Thi& -o,tion!"it%- i& 4h!t i& 0ehin) the con$e/it% o( e&e! ch o'tcome&* An o,tion !""o4& it& '&e to get mo e ',&i)e th!n )o4n&i)e !& he c!n &e"ect !mong the e&'"t& 4h!t (it& him !n) (o get !0o't the e&t (he h!& the o,tion# not the o0"ig!tion)* 6ence o' 'n)e &t!n)ing o( o,tion!"it% c!n 0e e/ten)e) to e&e! ch , og !m& 7 thi& )i&c'&&ion i& moti$!te) 0% the (!ct th!t the !'tho &,ent mo&t o( hi& !)'"t "i(e !& !n o,tion t !)e * I( 4e t !n&"!te F !n=oi& >!co0;& i)e! into the&e te m&# e$o"'tion i& ! con$e/ ('nction o( &t e&&o & !n) e o & 7genetic m't!tion& come !t no co&t !n) ! e et!ine) on"% i( the% ! e !n im, o$ement* So ! e the !nce&t !" he' i&tic& !n) '"e& o( th'm0& em0e))e) in &ociet%2 (o me) "i+e eci,e& 0% contin'o'&"% t!+ing the ',,e .0o'n) o( -4h!t 4o +&-* B't 'n"i+e n!t' e 4he e choice& ! e m!)e in !n !'tom!tic 4!% $i! &' $i$!"# h'm!n o,tion!"it% e8'i e& the e/e ci&e o( !tion!" choice to !tchet ', to &omething 0ette th!n 4h!t , ece)e& it 7!n)# !"!&# h'm!n& h!$e ment!" 0i!&e& !n) c'"t' !" hin) !nce& th!t n!t' e )oe&n;t h!$e* O,tion!"it% ( ee& '& ( om the &t !ight?!c+et o( )i ection# , e)iction&# ,"!n&# !n) n! !ti$e&* (To '&e ! met!,ho ( om in(o m!tion theo %# i( %o' ! e going to ! $!c!tion e&o t o((e ing %o' mo e o,tion&# %o' c!n , e)ict %o' !cti$itie& 0% !&+ing ! &m!""e n'm0e o( 8'e&tion& !he!) o( time*) <hi"e getting ! 0ette eci,e (o h'mm'& 4i"" not ch!nge the 4o ")# &ome e&'"t& o((e !0no m!""% "! ge 0ene(it& ( om )i&co$e %2 con&i)e ,enici""in o chemothe !,% o ,otenti!" c"e!n techno"ogie& !n) &imi"! high im,!ct e$ent& (-B"!c+ S4!n&-)* The )i&co$e % o( the (i &t !ntimic o0i!" ) 'g& c!me !t the hee" o( h'n) e)& o( &%&tem!tic (con$e/) t i!"& in the 9@AB& 0% &'ch ,eo,"e !& Dom!g+ 4ho&e e&e! ch , og !m con&i&te) in t %ing o't )%e& 4itho't m'ch 'n)e &t!n)ing o( the 0io"ogic!" , oce&& 0ehin) the e&'"t&* An) 'n"i+e !n e/,"icit (in!nci!" o,tion (o 4hich the 0'%e ,!%& ! (ee to ! &e""e # hence ten) to t !)e in ! 4!% to , e$ent 'n)'e , o(it&# 0ene(it& ( om e&e! ch ! e not 3e o.&'m* T6INGS LOVE UNCERTAINTY <h!t !""o4& '& to m!, ! e&e! ch ('n)ing !n) in$e&tment metho)o"og% i& ! co""ection o( m!them!tic!" , o,e tie& th!t 4e h!$e +no4n he' i&tic!""% &ince !t "e!&t the 9CBB& !n) e/,"icit"% &ince ! o'n) 9@BB (4ith the e&'"t& o( >oh!n >en&en !n) Lo'i& B!che"ie )* The&e , o,e tie& i)enti(% the ine$it!0i"it% o( g!in& ( om con$e/it% !n) the co'nte int'iti$e 0ene(it o( 'nce t!int% ii iii* Let '& c!"" the -con$e/it% 0i!&the )i((e ence 0et4een the e&'"t& o( t i!" !n) e o in 4hich g!in& !n) h! m ! e e8'!" ("ine! )# !n) one in 4hich g!in& !n) h! m ! e !&%mmet ic ( to e,e!t# ! con$e/ ,!%o(( ('nction)* The cent !" !n) '&e('" , o,e tie& ! e th!t !) The mo e con$e/ the ,!%o(( ('nction# e/, e&&e) in )i((e ence 0et4een ,otenti!" 0ene(it& !n) h! m# the "! ge the 0i!&* 0) The mo e $o"!ti"e the en$i onment# the "! ge the 0i!&* Thi& "!&t , o,e t% i& mi&&e) !& h'm!n& h!$e ! , o,en&it% to h!te 'nce t!int%* #ntifragile i& the n!me thi& !'tho g!$e ((o "!c+ o( ! 0ette one) to the 0 o!) c"!&& o( ,henomen! en)o4e) 4ith &'ch ! con$e/it% 0i!&# !& the% g!in ( om the -)i&o )e c"'&te -# n!me"% $o"!ti"it%# 'nce t!int%# )i&t' 0!nce&# !n)omne&&# !n) &t e&&o &* The !nti( !gi"e i& the e/!ct o,,o&ite o( the ( !gi"e 4hich c!n 0e )e(ine) !& h!ting )i&o )e * A co((ee c', i& ( !gi"e 0ec!'&e it 4!nt& t !n8'i"it% !n) ! "o4 $o"!ti"it% en$i onment# the !nti( !gi"e 4!nt& the o,,o&ite1 high $o"!ti"it% inc e!&e& it& 4e"(! e* Thi& "!tte !tt i0'te# g!ining ( om 'nce t!int%# (!$o & o,tion!"it% o$e the te"eo"ogic!" in !n o,!8'e

&%&tem# !& it c!n 0e &ho4n th!t the te"eo"ogic!" i& h' t 'n)e inc e!&e) 'nce t!int%* The ,oint c!n 0e m!)e c"e! 4ith the (o""o4ing* <hen %o' in?ect 'nce t!int% !n) e o & into !i ,"!ne i)e (the ( !gi"e o conc!$e c!&e) the e&'"t i& 4o &ene)# !& e o & in$! i!0"% "e!) to ,"!ne )e"!%& !n) inc e!&e) co&t& 7 not co'nting ! ,otenti!" ,"!ne c !&h* The &!me 4ith 0!n+ ,o t(o"io& !n) ( !gi"e con&t 'ct&* B't it %o' in?ect 'nce t!int% into ! con$e/ e/,o&' e &'ch !& &ome t%,e& o( e&e! ch# the e&'"t im, o$e&# &ince 'nce t!int% inc e!&e& the ',&i)e 0't not the )o4n&i)e* Thi& )i((e enti!" m!,& the 4!%* The con$e/it% 0i!&# 'n"i+e &e en)i,it% et !"*# c!n 0e )e(ine)# (o m!"i3e)# i)enti(ie)# e$en on the occ!&ion me!&' e) &cienti(ic!""%# !n) c!n "e!) to ! (o m!" ,o"ic% o( )eci&ion m!+ing 'n)e 'nce t!int%# !n) c"!&&i(% &t !tegie& 0!&e) on thei e" ante , e)icte) e((icienc% !n) , o?ecte) &'cce&&# !& 4e 4i"" )o ne/t 4ith the (o""o4ing C '"e&*

Figure ( The #ntifragilit$ Edge &)onve"it$ *ias'. # random simulation shows the difference %etween a' the process with conve" trial and error &antifragile' %' a process of pure nowledge devoid of conve" tin ering & nowledge %ased', c' the process of nonconve" trial and error+ where errors are e,ual in harm and gains &pure chance'. #s we can see there are domains in which rational and conve" tin ering dwarfs the effect of pure nowledge
iv.

SEVEN RULES OF ANTIFRAGILITY (CONVEXITY) IN RESEARC6 Ne/t I o't"ine the '"e&* In ,! enthe&e& ! e (!ncie 4o )& th!t "in+ the i)e! to o,tion theo %* 9) Con$e/it% i& e!&ie to !tt!in th!n +no4"e)ge (in the technic!" ?! gon# the -"ong.g!mm!, o,e t%)1 A& 4e &!4 in Fig' e A# 'n)e &ome "e$e" o( 'nce t!int%# 4e 0ene(it mo e ( om im, o$ing the ,!%o(( ('nction th!n ( om +no4"e)ge !0o't 4h!t e/!ct"% 4e ! e "oo+ing (o * Con$e/it% c!n 0e inc e!&e) 0% "o4e ing co&t& ,e 'nit o( t i!" (to im, o$e the )o4n&i)e)* A) A -9DN- &t !teg% i& !"mo&t !"4!%& 0e&t 4ith con$e/ &t !tegie& (the )i&,e &ion , o,e t%)1(o""o4ing ,oint (9) !n) e)'cing the co&t& ,e !ttem,t# com,en&!te 0% m'"ti,"%ing the n'm0e o( t i!"& !n) !""oc!ting 9DN o( the ,otenti!" in$e&tment !c o&& N in$e&tment&# !n) m!+e N !& "! ge !& ,o&&i0"e* Thi& !""o4& '& to minimi3e the , o0!0i"it% o( mi&&ing !the th!n m!/imi3e , o(it& &ho'") one h!$e ! 4in# !& the "!tte te"eo"ogic!" &t !teg% "o4e & the , o0!0i"it% o( ! 4in* A "! ge e/,o&' e to ! &ing"e t i!" h!& "o4e e/,ecte) et' n th!n ! ,o t(o"io o( &m!"" t i!"&*

F' the # e&e! ch ,!%o((& h!$e -(!t t!i"&-# 4ith e&'"t& in the -t!i"&- o( the )i&t i0'tion )omin!ting the , o,e tie&2 the 0'"+ o( the g!in& come ( om the ! e e$ent# -B"!c+ S4!n-1 9 in 9BBB t i!"& c!n "e!) to EBF o( the tot!" cont i0'tion&7&imi"! to &i3e o( com,!nie& (EBF o( c!,it!"i3!tion o(ten come& ( om 9 in 9BBB com,!nie&)# 0e&t&e""e & (thin+ 6! % Potte )# o 4e!"th* An) c itic!""% 4e )on;t +no4 the 4inne !he!) o( time*

Figure --Fat Tails. /mall Pro%a%ilit$, 0igh 1mpact Pa$offs. The hori2ontal line can %e the pa$off over time, or cross-sectional over man$ simultaneous trials.

G) Se i!" o,tion!"it% (the c"i8'et , o,e t%)* A igi) 0'&ine&& ,"!n get& one "oc+e) into ! , e&et in$! i!nt ,o"ic%# "i+e ! high4!% 4itho't e/it& 7hence )e$oi) o( o,tion!"it%* One nee)& the !0i"it% to ch!nge o,,o t'ni&tic!""% !n) - e&et- the o,tion (o ! ne4 o,tion# 0% !tcheting ',# !n) getting "oc+e) ', in ! highe &t!te* To t !n&"!te into , !ctic!" te m&# ,"!n& nee) to 9) &t!% ("e/i0"e 4ith ( e8'ent 4!%& o't# !n)# co'nte to int'ition A) 0e $e % &ho t te m# in o )e to , o,e "% c!,t' e the "ong te m* 5!them!tic!""%# (i$e &e8'enti!" one.%e! o,tion& ! e $!&t"% mo e $!"'!0"e th!n ! &ing"e (i$e.%e! o,tion* Thi& e/,"!in& 4h% m!tte & &'ch !& &t !tegic ,"!nning h!$e ne$e 0o n ( 'it in em,i ic!" e!"it%1 ,"!nning h!& ! &i)e e((ect to e&t ict o,tion!"it%* It !"&o e/,"!in& 4h% to,.)o4n cent !"i3e) )eci&ion& ten) to (!i"* H) Nonn! !ti$e Re&e! ch (the o,tion!"it% , o,e t%)* Techno"ogi&t& in C!"i(o ni! -h! $e&ting B"!c+ S4!n&- ten) to in$e&t 4ith !gent& !the th!n ,"!n& !n) n! !ti$e& th!t "oo+ goo) on ,!,e # !n) !gent& 4ho +no4 ho4 to '&e the o,tion 0% o,,o t'ni&tic!""% &4itching !n) !tcheting ', 7t%,ic!""% ,eo,"e t % &i/ o &e$en techno"ogic!" $ent' e& 0e(o e getting to )e&tin!tion* Note the (!i"' e in -&t !tegic ,"!nning- to com,ete 4ith con$e/it%* E) Theo % i& 0o n ( om (con$e/) , !ctice mo e o(ten th!n the e$e &e (the nonte"eo"ogic!" , o,e t%)* Te/t0oo+& ten) to &ho4 techno"og% ("o4ing ( om &cience# 4hen it i& mo e o(ten the o,,o&ite c!&e# )'00e) the -"ect' ing 0i )& on ho4 to ("%- e((ect $ $i* In &'ch )e$e"o,ment& !& the in)'&t i!" e$o"'tion (!n) mo e gene !""% o't&i)e "ine! )om!in& &'ch !& ,h%&ic&)# the e i& $e % "itt"e hi&to ic!" e$i)ence (o the cont i0'tion o( ('n)!ment!" e&e! ch com,! e) to th!t o( tin+e ing 0% ho00%i&t&* $ii Fig' e A &ho4&# mo e technic!""% ho4 in ! !n)om , oce&& ch! !cte i3e) 0% -&+i""&- !n) -"'c+-# !n) &ome o,!cit%# !nti( !gi"it% 7the con$e/it% 0i!&7 c!n 0e &ho4n to &e$e e"% o't,e (o m -&+i""&-* An) con$e/it% i& mi&&e) in hi&to ie& o( techno"ogie&# e,"!ce) 4ith e/ ,o&t n! !ti$e&* I) P emi'm (o &im,"icit% (the "e&&.i&.mo e , o,e t%)* It too+ !t "e!&t (i$e mi""enni! 0et4een the in$ention o( the 4hee" !n) the inno$!tion o( ,'tting 4hee"& 'n)e &'itc!&e&* It i& &ometime& the

&im,"e&t techno"ogie& th!t ! e igno e)* In , !ctice the e i& no , emi'm (o com,"e/i(ic!tion2 in !c!)emi! the e i&* Loo+ing (o !tion!"i3!tion&# n! !ti$e& !n) theo ie& in$ite& (o com,"e/it%* In !n o,!8'e o,e !tion to (ig' e o't e/ !nte 4h!t +no4"e)ge i& e8'i e) to n!$ig!te i& im,o&&i0"e* C) Bette c!t!"og'ing o( neg!ti$e e&'"t& (the $i! neg!ti$! , o,e t%)* O,tion!"it% 4o +& 0% neg!ti$e in(o m!tion# e)'cing the &,!ce o( 4h!t 4e )o 0% +no4"e)ge o( 4h!t )oe& not 4o +* Fo th!t 4e nee) to ,!% (o neg!ti$e e&'"t&* Some o( the c itic& o( the&e i)e!& 7o$e the ,!&t t4o )ec!)e&7 h!$e 0een co'nte ing th!t thi& , o,o&!" e&em0"e& 0'%ing -"otte % tic+et&-* Lotte % tic+et& ! e ,!tent"% o$e , ice)# e("ecting the -"ong &hot 0i!&- 0% 4hich !gent&# !cco )ing to economi&t&# o$e ,!% (o "ong o))&* Thi& com,! i&on# it t' n& o't i& (!""!cio'&# !& the e((ect o( the "ong &hot 0i!& i& "imite) to ! ti(ici!" &et',&1 "otte ie& ! e &te i"i3e) !n)omne&&# con&t 'cte) !n) &o") 0% h'm!n&# !n) h!$e ! +no4n ',,e 0o'n)* Thi& !'tho c!""& &'ch ! , o0"em the -"')ic (!""!c%-* Re&e! ch h!& e/,"o&i$e ,!%o((&# 4ith 'n+no4n ',,e 0o'n) 7! -( ee o,tion-# "ite !""%* An) 4e h!$e e$i)ence (( om the ,e (o m!nce o( 0!n+&) th!t in the e!" 4o ")# 0etting !g!in&t "ong &hot& )oe& not ,!%# 4hich m!+e& e&e! ch ! (o m o( e$e &e.0!n+ing$ iii *