COMMENT ⊗   VALID 00009 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00003 00002	.DEVICE XGP
C00004 00003	.PORTION TITLEPAGE
C00005 00004	.PAGE FRAME 48 HIGH 89 WIDE
C00006 00005	2RECOGNITION GROUPING*
C00015 00006	2CHANGES*		Is this rele. to producing the desired change in the world?
C00017 00007	2FINAL*		What situations is this β rele. to bringing about?
C00019 00008	2PAST*		Where is this used frequently, to advantage?
C00021 00009	2IDEN* {not}{quick}  	  {fast} tests to see if this β is {not} currently referred to
C00023 ENDMK
C⊗;
.DEVICE XGP
.!XGPCOMMANDS←"/TMAR=30/PMAR=2130/BMAR=40"

.FONT 1 "BASL30"
.FONT 2 "BDR40"
.FONT 4  "BASI30"
.FONT 5 "BASB30"
.FONT 7  "NGR20"
.TURN ON "↑↓_π{"
.TURN ON "⊗" FOR "%"
.PAGE FRAME 50 HIGH 89 WIDE
.AREA TEXT LINES 4 TO 48
.AREA HEADING LINES 1 TO 3
.AREA FOOTING LINE 50
.!XGPLFTMAR←160
.SPACING 55 MILLS
.PREFACE 160 MILLS
.NOFILL
.PREFACE 45 MILLS
.FILL
.COUNT PAGE PRINTING "1"
.PAGE←0
.NEXT PAGE
.MACRO B ⊂ BEGIN VERBATIM GROUP ⊃
.MACRO E ⊂ APART END ⊃
.TABBREAK
.PORTION TITLEPAGE
.GROUP SKIP 5
.BEGIN CENTER RETAIN
⊗2↓_GIVEN KNOWLEDGE_↓⊗*

FOR A SYSTEM WHICH WILL DO   MATHEMATICAL  THEORY  FORMATION⊗*


⊗7A  System  which  can  develop  mathematical  concepts  intuitively⊗*
.END
.GROUP SKIP 8
.NOFILL
⊗5Doug Lenat

Avra Cohn


Stanford University
Artificial Intelligence Laboratory

⊗*


THIRD  SKETCH   of initial knowledge

⊗4Not for distribution⊗*




.ONCE CENTER
⊗7{DATE}⊗*
.NEXT PAGE
.PAGE FRAME 48 HIGH 89 WIDE
.AREA TEXT LINES 3 TO 47
.AREA HEADING LINE 1
.AREA FOOTING LINE 48
.EVERY HEADING(⊗7GIVEN KNOWLEDGE    {DATE}⊗*, ⊗4Doug Lenat and Avra Cohn⊗*,⊗7page {PAGE}⊗*)
.!XGPLFTMAR←120
.SPACING 55 MILLS
.PREFACE 160 MILLS
.NOFILL
.PREFACE 17 MILLS
.PAGE←0
.NEXT PAGE
.NOFILL
.TURN OFF "{"
.INDENT 0

⊗2RECOGNITION GROUPING⊗*



⊗5RECOGNITION GROUPING⊗*  How to recognize when the relevant task is to determine rele.
 CHANGES  {(rele. part  goes: unknown to known, .80, .04, intu: piece = rele. part)}
 FINAL 	{(the rele. gp. is Recog), .99, .03, intu: puz. piece wanted is in Recog packet)}
 PAST
 IDEN {not}{quick}


⊗5ACT GROUPING⊗*  How to deal with the RECOG group of parts of a BEING named β.
 FILLIN  
 STRUCTURE  Never split. Might restructure if heavily used and very many entries.
 CHECK 
 REPRESENTATION 

 
⊗5INFO GROUPING⊗*
 DEFINITION
 INTU  Several puzzle pieces are needed; β is a packet of pieces; RECOG says when to try them all.
 TIES  Up: Archtype BEING.
.SKIP TO COLUMN 1
⊗2CHANGES⊗*		Is this rele. to producing the desired change in the world?


⊗5RECOGNITION GROUPING⊗*  How to recognize when the relevant part to work on is Changes.
 CHANGES
 FINAL 	{(the rele. part is Changes), .99, .03, intu: puzzle piece turns out to be Changes)}
 PAST
 IDEN {not}{quick}


⊗5ACT GROUPING⊗*  How to deal with the CHANGES part of a BEING named β.
 FILLIN  The first few times β is succ. used, carefully monitor ⊗4all⊗* changes. 
	Comb through, each time β is reused, updating probs. for each listed change.
	Elim. any entry if its prob. is low and it can't be justified quickly.
	Watch especially for Changes mentioned in BEINGs listed in Ties part of β.
 STRUCTURE  Partition based on types of changes, on prob., and/or on centrality to β.
 CHECK   Each change should be rare overall compared to its prob. 
	No prob. should be < 5%. 
	Any change with prob. < 30% should be justifiable (intui. and/or formal).
 REPRESENTATION  (change in world, centrality, probability, justification)*

 
⊗5INFO GROUPING⊗*
 DEFINITION  The change component reflects a difference in the world which β is rele. to.
 INTU
 TIES  Up: Recog. grouping.
.SKIP TO COLUMN 1
⊗2FINAL⊗*		What situations is this β rele. to bringing about?


⊗5RECOGNITION GROUPING⊗*  How to recognize when the relevant part to work on is Final.
 CHANGES
 FINAL 	{(the rele. part is Final), .99, .03, intu: puzzle piece turns out to be Final)}
 PAST
 IDEN {not}{quick}


⊗5ACT GROUPING⊗*  How to deal with the FINAL part of a BEING named β.
 FILLIN  The first few times β is succ. used, carefully monitor ⊗4all⊗* products.
	Comb through, each time β is reused, updating probs. for each listed statement.
	Elim. any entry if its prob. is low and it can't be justified quickly.
	Watch especially for Final products mentioned in BEINGs listed in β.Ties.
 STRUCTURE  Partition based on types of products, on prob., and/or on centrality to β.
 CHECK   Each final outcome should be rare overall compared to its prob. 
	No prob. should be < 5%. 
	Any entry with prob. < 30% should be justifiable (intui. and/or formal).
 REPRESENTATION  (future feature or byproduct, centrality, probability, justification)*

 
⊗5INFO GROUPING⊗*
 DEFINITION  The Final product component reflects a feature true in the near future,rele. to β.
 INTU
 TIES  Up: Recog. grouping.
.SKIP TO COLUMN 1
⊗2PAST⊗*		Where is this used frequently, to advantage?


⊗5RECOGNITION GROUPING⊗*  How to recognize when the relevant part to work on is Past.
 CHANGES
 FINAL 	{(the rele. part is Past), .99, .03, intu: puzzle piece turns out to be Past)}
 PAST
 IDEN {not}{quick}


⊗5ACT GROUPING⊗*  How to deal with the PAST part of a BEING named β.
 FILLIN  The first few times β is succ. used, extract all int. features of initial sit.
	When reused, update probs. Remove freak entries. 
	Watch especially for features mentioned in Past part of BEINGs listed in β.Ties.
 STRUCTURE  Partition based on types of features, on prob., and/or on centrality to β.
 CHECK   Each situation feature should be rare overall compared to its prob. 
	No prob. should be < 4%. 
	Any entry with prob. < 20% should be justifiable (intui. and/or formal).
 REPRESENTATION  (current feature of world, centrality, probability, justification)*

 
⊗5INFO GROUPING⊗*
 DEFINITION  This component makes predictions based on past trends; inference of β's uses.
 INTU  If it worked once, it'll work again in a similar situation. Treatment as black box.
 TIES  Up: Recog. grouping. 
.SKIP TO COLUMN 1
⊗2IDEN⊗* {not}{quick}  	  {fast} tests to see if this β is {not} currently referred to


⊗5RECOGNITION GROUPING⊗*  How to recognize when the relevant part to work on is Iden.
 CHANGES
 FINAL 	{(the rele. part is Iden), .99, .03, intu: puzzle piece turns out to be Iden)}
 PAST
 IDEN {not}{quick}


⊗5ACT GROUPING⊗*  How to deal with the IDEN part of a BEING named β.
 FILLIN  Initially, with low priority, ask the user how he means to refer to β.
	Whenever β is found relevant, ensure that it and its variations are included.
	Imp. to distinguish phrases for β from those for all the BEINGs in β.Ties.
 STRUCTURE  Partition based on syn, variat.,freq. of referral, user-name, centrality.
 CHECK   Each phrase should be rare overall compared to its prob. 
	No prob. should be < 20%. 
	Any entry with prob. < 50% should be justifiable (intui. and/or formal).
 REPRESENTATION  (phrase or word, centrality, probability, justification)*

 
⊗5INFO GROUPING⊗*
 DEFINITION   Phrase is English referral to β, esp. by User.
 INTU 
 TIES  Up: Recog. grouping. 
.SKIP TO COLUMN 1