Revised pats to e746

Protocol for Organization of the AIM Community

SUMEX Qroperel — Wrnunle 14,0973

Wwe attempt to reconcile the divergent impulses of sustaining
(1) the accountability of the AI tacility ho a fair response
to national needs, and (2) the responsibility and authority of
the principal investigator for the intellectual productivity
of his local program and of a community of close
collaborators.

ve

In part we have cespdonded to these concerns specifying a
Aiurston
prenegotiated allocatiom of the SUMEX tacility 40 the local
DENDRAL project and te AIM respectively. As discussed
already, the 50% allocation is’ based (B on the initial
configuration, (2) includes latitude for system effort and for
informal and preliminary associations, and (3) would be
reevaluated (i.e. negotiated between ALM and DENDRAL) at the
point of any augmentation of the facility. We do not imply
that DENPRAL would have an automatic claim to 5u% or any other
moiety of increases justified in terms of demands from the AI&
community. Alternatively, DENDRAL should be allowed to
comp@te without prejudice, for such incremental facilities.
View)
The National Committee (NCATM) should have the tinal

authority for the selection of AIM users and for the
disposition of the SUMEK resources designated for that
community. This includes not only the machine cycles but also
a prorata part of the personnel effort listed on Page 92 (Many

en em A0 AOE BRR een amsamneamae,

of the listed individuals serveg at a systen level, ythe needs

of both sectors} )

The Principal Investigator is responsible to the NCAIM for
rulfilling the contractual obligations of sharing the SUMEX
resource. He will report regularly to Carn, aad-thia in
tern will report to BkB concerning the adequacy of
fulfillment. BRB —teee has the right and responsibiity to
investigate, even interrupt the funding of the SUMEX resource
in the event of an adverse report; indeed it would have the
right to recover damages (as it can with other grants) in the
evert of flagrant violations of the grant instrument as

accepted by the University.

The deterrence of such an adverse report is a powerful and
obvious incentive to assure compliance with mutually agreed

objectives.

In discharging his responsibility the P.1. will appoint a
Facility Director ang a Liaison Coordinator who are instructed
to comply with these policies and to maintain close
communications with the users and with NCAIM. the P.I. will
not be a passive instrumert but will take an active role in

recruiting candidate users for admission to the AIM community,
and in sustaining a high level of communication with and

among ther.

Major system software policies will be established by the

NCAIM keeping in mind not only the immediate impact on the

DENDRAL and AIM sectors, but their overall effect on the

general development of AIL research. Likewise, the AIM effort

will seek to recruit other centers to cooperate (without yen arr
waking long range commitments like SUMEX) in these software

efforts over the net, and to share ad hoc backup and other

Machine resources.

We believe that many more potential users can be found thapy

wo have identified as yet -- and indeed who will reveal and

detend their projects pri the authentication that a

service will be provided. We will not, therefore, preselect

the user group until the facility has been advertised and

proposals have been formulated, although we have identified

some promising candidates. Some of the conflict-of-interest
difficulties that will attend this process are obvious, and

are an important reason for leaving the final selection to mawll
wedefined process that itself has been reviewed by the Study

Section and the NARRC.

The criteria for user selection might be ad hoc, relating
essentially to the originality and competence of the proposer,

and the relevance of his proposal to the general mission of
A.I. in Medicine. Indeed, this parallels what study sections
generally mast now do which comprises much, more judgement than

algorithm.

However, the mutual compatibility or rather reinforcement of
the efforts of the prospective users should be an important
criterion. Compatibility may be needed for the most efficient
cooperation, e.g. at the level of languages; complementarity
will add to the overall power of the repertoire of the
subprograms and conceptual approaches available to the group.
In addition, a number of other low-intensity and short-term
users might also ke coopted who simply promise a high return
of useful research in proportion to the resources (computer
cycles, administrative, and intellectual attention) that they

consume.

AX
The initial user Jour should be selected by a 7-member OCA, on
Organizing Committee whose membership can be designated as

fellows:

2 Members -- to be named by BRB

Alan Newell -~ past member of the study section
Selfridge-- past member of the study section
Langridye-- past member of the study section
Feigenkaum

Amarel
The OCAIM will advertise the facility (with
support from SUMEX) and review proposals 90
interval will be needed anyhow to check out
service to a reasonable levelfof utility to
that time, OCAIM will designate the initial

ground rules of their relationship to SUMEX,

active staff

days later®. This
the network

the projects. At
user group, the

the initial

wound for NCAIM, and the initial merkership of NCAIM. The

ongoing committee, NCAIM, may now include users whose role in

AIM has been previously reviewed by the organizing committee.

We propose that NCAIM include six members:

1 - representing SUMEX

3 - representing other users

2 - from #CAIM or other non-user personnel

(The details of these numbers are of course negotiabfle.) We

need a committee large enough to retain a representative

quorum despite occasicnal absences, small enough to meet often

(6 to 8 times per year) and decisively.

NCAIM would rotate its membership, replacing

by nominating a slate for approval by HRB.

two each year,

Resides continuing the processor user so1cttion and f

monitoring service standards and accountability of SUMEX,

NCATM would also estatlish other working relationships at

various levels to advise SUMEX management on

its programs of
Systems and language support, user education, documentation,
and conferences involving larger components of the computer

science and medical research communities. ne 4

we believe that NCAIN will peempety recruit a considerable

gtoup cf usersusto cr} seg. Comreeme The ovrorcfayts
aro . NEM pitt Can herr He Lot

wearony wat Ce BS. Im plaworey nr oligo read v
Catan te pect diffeuckinted sem gerop. Brat iiahy

shen of Hbigenied Reigns on Q1r.2_ yt Ce

a AAR Warrose yet of Oe Arata Labor a fres
yn pee Vier, Mhirg KX Ogos Beha 41 0 L020

Wow Ais (Oa Aaralll rrp pe De oe
Away be e yoMe pond evaeye orn) Hu unk.

Lardas) ~f ia Hoe leagues orrd Q) Le seok o te

Oelies S 4 ee tts tlt ch
SE fom sortase a at. fe Ab frr$,
— culos conte ALE be ALY ofa Lokin steye,

BT coy rrget iu SUA ot Dh part Coon

q sth a
$2)
de
Ge