Toward a Critical Technical Practice

tags

Critical Technical Practice Philip E. Agre

Notes

no idea how to begin reckoning their effects upon society, assuming that such a global notion of "effects" even makes sense.

NOTER_PAGE: (1 . 0.538107752956636)

computing has been constituted as a kind of imperialism; it aims to reinvent virtually every other site of practice in its own image.

NOTER_PAGE: (1 . 0.6984231274638634)

translator between languages and worldviews: the formalisms of computing and the craft culture of the "application domain".

NOTER_PAGE: (2 . 0.09264126149802891)

AI appealed to my adolescent sensibilities

NOTER_PAGE: (2 . 0.19448094612352168)

Criticisms of the field, no matter how sophisticated and scholarly they might be, are certain to be met with the assertion that the author simply fails to understand a basic point.

NOTER_PAGE: (2 . 0.3436268068331143)

reform the field by providing it with the critical methods it needed – a critical technical practice.

NOTER_PAGE: (2 . 0.41458607095926414)

Without the idea that ideologies and social structures can be reproduced through a myriad of unconscious mechanisms such as linguistic forms and bodily habits, all critical analysis may seem like accusations of conscious malfeasance.

NOTER_PAGE: (2 . 0.8252299605781866)

counterrevolution against behaviorism – a way to make talk about mental processes scientific and precise.

NOTER_PAGE: (3 . 0.39290407358738505)

MIT in particular came to prominence as a technical university that had made significant innovations in the war effort

NOTER_PAGE: (3 . 0.44415243101182655)

the "closed world" – the whole world as one large technical system

NOTER_PAGE: (3 . 0.49737187910643893)

AI was one part of an overall movement with a coherent worldview

NOTER_PAGE: (3 . 0.6051248357424441)

Few of them regarded themselves as engineers seeking purely instrumental solutions to technical problems.

NOTER_PAGE: (3 . 0.8948751642575559)

a cult of "smartness", such that students deemed "smart" (including most everyone accepted to graduate school) were given wide latitude to pursue their own research directions, no matter how odd

NOTER_PAGE: (4 . 0.49408672798948755)

If the field of AI during those decades was a servant of the military then it enjoyed a wildly indulgent master.

NOTER_PAGE: (4 . 0.5249671484888305)

reached back three hundred years to identify themselves with the philosophy of Descartes

NOTER_PAGE: (4 . 0.7063074901445466)

philosophical basis of functionalism (Fodor 1968), provided the pioneers of AI with a vocabulary through which rule-based accounts of cognitive rationality could be rendered mechanical while also being meaningfully treated as mental phenomena

NOTER_PAGE: (4 . 0.9067017082785809)

The central practice of the field, and its central value, was technical formalization.

NOTER_PAGE: (5 . 0.10052562417871222)

tended to treat their research as the heir of virtually the whole of intellectual history

NOTER_PAGE: (5 . 0.13994743758212877)

often heard AI people portray philosophy, for example, as a failed project, and describe the social sciences as intellectually sterile.

NOTER_PAGE: (5 . 0.1583442838370565)

The early demonstrations of AI were incomparably more sophisticated than the mechanistic philosophies of even a short time before. As a result, the people who had stayed up many late nights getting them to work felt justified in extrapolating this extraordinary rate of progress for one or two or three more decades at least.

NOTER_PAGE: (5 . 0.3002628120893561)

use of computer symbols such as REASON and DECIDE and GOAL whose relationship to the actual human phenomena that those words ordinarily name is suggestive at best

NOTER_PAGE: (5 . 0.37582128777923784)

paradoxically, an intensely social culture of seemingly quite asocial people

NOTER_PAGE: (5 . 0.8101182654402103)

opposed the falseness of bureaucratic life to the principled meritocracy of their craft

NOTER_PAGE: (5 . 0.8304862023653088)

the "work ethic": it has to work. The "result" of an AI research project is a working system

NOTER_PAGE: (5 . 0.8804204993429698)

The field, accordingly, reckons its history primarily as a sequence of computer systems and secondarily as a history of debates among different approaches to the construction of systems.

NOTER_PAGE: (6 . 0.07030223390275953)

research programming in AI is self- consciously virtuosic and experimental. Much of the field's internal discourse has been concerned with the intuitions that guide the design of its complex, ambitious systems.

NOTER_PAGE: (6 . 0.21484888304862024)

The point is certainly not that AI people are mystics, or that they consciously wish to make anything obscure, but rather that they take seriously the craft nature of their work.

NOTER_PAGE: (6 . 0.30486202365308807)

The enormous obstinacy of technical work – if a method cannot be made to work in a given case then no amount of sloppiness or vagueness will make it work – seems to back these potentially nebulous intuitions with a "hardness" and irrefutability

NOTER_PAGE: (6 . 0.4875164257555848)

What matters practically is not the vague issue of what the words "really mean" but the seemingly precise issue of how they can be defined in formal terms that permit suitably narratable systems to be designed.

NOTER_PAGE: (6 . 0.7542706964520368)

This dual character of AI terminology – the vernacular and formal faces that each technical term presents – has enormous consequences

NOTER_PAGE: (7 . 0.24967148488830487)

finally one reaches a suitably elementary repertoire of actions from which all others are assembled. Miller, Galanter, and Pribram never offered a definitive set of these primitive actions, and the field has never felt it necessary to do so.

NOTER_PAGE: (8 . 0.053219448094612355)

The idea that human beings do not conduct their lives by means of planning, for example, is just short of unintelligible.

NOTER_PAGE: (8 . 0.3226018396846255)

The term "planning", in other words, exhibits an elastic quality: as a technical proposition it refers to a quite specific and circumscribed set of functionalities and algorithms, but as an empirical proposition it refers to anything at all that can plausibly be glossed with the term.

NOTER_PAGE: (8 . 0.3587385019710907)

attempts to deny the narrow technical theory sound to the ears of AI researchers like denials that the sequential ordering of human behavior is determined by any coherent process at all.

NOTER_PAGE: (8 . 0.599868593955322)

the tendency to conflate representations with the things that they represent.

NOTER_PAGE: (8 . 0.6668856767411301)

Many early AI researchers were clearly attempting, at one level or another, to reproduce their own psyches on computers, and many of them drew on introspection to motivate their programs. Introspection as a formal research method in psychology, of course, had been comprehensively discredited decades earlier. But AI people have not regarded introspection as evidence but as inspiration

NOTER_PAGE: (10 . 0.3587385019710907)

I became interested in what I called "hassles", which are small bits of trouble

NOTER_PAGE: (10 . 0.6491458607095927)

my dissertation (Agre 1988), once I finally wrote it, was motivated largely by a passion to explain to my fellow AI people how our AI concepts had cut us off from an authentic experience of our own lives. I still believe this.

NOTER_PAGE: (11 . 0.053219448094612355)

Many technical people have observed that phenomenological texts, when read as specifications for technical mechanisms, sound like mysticism. This is because Western mysticism, since the great spiritual forgetting of the later Renaissance, is precisely a variety of mechanism that posits impossible mechanisms

NOTER_PAGE: (11 . 0.4855453350854139)

Their stance was: if your alternative is so good then you will use it to write programs that solve problems better than anybody else's, and then everybody will believe you. Even though I believe that building things is an important way of learning about the world, nonetheless I knew that this stance was wrong, even if I did not understand how.

NOTER_PAGE: (12 . 0.5913272010512484)

many valuable criticisms will go unpublished if all research papers are required to present new working systems

NOTER_PAGE: (12 . 0.7555847568988173)

AI people often make substantive claims about knowledge or learning or language, and yet many of them will respond with indignation to arguments that their projects fundamentally misconstrue the nature of these phenomena; in most cases (the primary exception being Newell and Simon's research group at Carnegie-Mellon University) they will argue not that the claims against their work are empirically false but that they are non sequiturs.

NOTER_PAGE: (12 . 0.9388961892247043)

The seemingly commonsensical demand to prove alternatives in practice is thus actually a demand to express disagreements with the existing language within the existing language itself, and this is nearly impossible.

NOTER_PAGE: (13 . 0.3968462549277267)

AI's elastic use of language ensures that nothing will seem genuinely new, even if it actually is, while AI's intricate and largely unconscious cultural system ensures that all innovations, no matter how radical the intentions that motivated them, will turn out to be enmeshed with traditional assumptions and practices.

NOTER_PAGE: (13 . 0.6865965834428384)

Orthographic projections are ubiquitous in the diagrams of AI papers; they make it seem reasonable that the simulated agent maintains a panoptic representation of its environment. This is one particularly insidious manifestation of AI's tendency to conflate representations and things-represented.

NOTER_PAGE: (14 . 0.19645203679369252)

the blocks have their names ("A", "B", "C", etc) written on them, as if the agent's mental symbols were part of the material world, automatically connected to the things they name.

NOTER_PAGE: (14 . 0.2542706964520368)

it is difficult to become aware of the full range of assumptions underneath existing practices

NOTER_PAGE: (14 . 0.7601839684625493)

having formulated an alternative intuition, it is difficult to reduce that intuition to a novel technical method

NOTER_PAGE: (14 . 0.775952693823916)

having invented a new technical method, it is difficult to prevent that method from being construed as "nothing new"

NOTER_PAGE: (14 . 0.8107752956636005)

having coupled a new technical method with a new way of talking about the phenomena, it is difficult to apply the method to any real cases without inventing a lot of additional methods as well

NOTER_PAGE: (14 . 0.8534822601839684)

Was our seemingly lonely work in the mid-1980's subliminally influenced by the ongoing changes in military thinking?

NOTER_PAGE: (15 . 0.21024967148488832)

Lucy Suchman

NOTER_PAGE: (15 . 0.5019710906701709)

When I first started trying to reform AI, I believed in revolutions. It seemed to me that I could clear the ground completely and start over, working out a whole alternative intellectual system that would replace everything that was there before.

NOTER_PAGE: (15 . 0.6241787122207622)

I believe in something more like hermeneutics. The intellectual utility of technical exercises, aside from the practical utility that they might actually have in the world, lies precisely in their limitations and failures.

NOTER_PAGE: (15 . 0.7496714848883049)

maintaining constructive engagement with researchers whose substantive commitments I found wildly mistaken

NOTER_PAGE: (16 . 0.12680683311432325)

nothing can substitute for the daily work of trying to get things built and working.

NOTER_PAGE: (16 . 0.45532194480946125)

A critical technical practice will, at least for the foreseeable future, require a split identity – one foot planted in the craft work of design and the other foot planted in the reflexive work of critique.

NOTER_PAGE: (16 . 0.5919842312746386)

forms of language, career strategies, and social networks that support the exploration of alternative work practices that will inevitably seem strange

NOTER_PAGE: (16 . 0.6872536136662286)