Data Epistemology and healthcare

There are some deep difficulties in blending data across healthcare functional and institutional boundaries. There are dozens of different reasons why data does not "meld" well, which as technology progresses and people move around more becomes a great hindrance.

With respect to payment processes, the U.S. government and others have imposed a very "coarse" view on an immensely complex reality. In the U.S., if you go to the hospital, your treatment will be force-fitted into one of a few hundred "diagnostic" groups, and the hospital will be compensated based on a payment level allotted to that group. All the rich detail of your experience is discarded for the sake of practicality, just as your passport photo may represent you with only a few hundred thousand pixels rather than with a portrait-quality pictures with many millions.

On the other hand, your actual diagnosis and treatment exhibits infinite particularity, and any system that is to aid in the healthcare process itself needs much more fidelity. Choices have to be made in terms of "fidelity" - e.g., although it may be possible to capture and retain in a database your every heartbeat, is there a need to do so? If we choose to capture only a very tiny subset of that data (e.g., your pulse as measured twice a day), are we depriving some data consumers of important knowledge? If we do capture great detail, is it possible that we will not associate it with you, but with some other patient? It happens.

Today, there is the general outcry that the healthcare system in the the U.S. is unnecessarily errorprone and expensive because of an inability to share data. Although one will rarely if ever associate that discussion with "epistemology," the barriers that we face are in fact our number one epistemological problem.

Owl versus Parrot

Owl or Parrot?

We originally and perhaps still think of our systems as environments that parrot back what we tell them. Indeed, this may be the best way to think of them, because in the ordinary sense of the words systems neither "think" nor "learn."

In many respects, systems that have to "comply" must indeed be faithful parrots. The system that is reporting financial results has to "parrot" back inputs again and again, with great consistency even though the individual outputs have varying specifications and purposes. Similarly "validated" systems used by pharmaceutical companies need to function as trustworthy parrots.

On the other hand, we have to deal with the reality that many of our systems are independently collecting masses of data from various sorts of sensors. Such systems perform various operations to that data - e.g., regulating its "fidelity" in terms of, say, bits per image - and with that data - where to store it, whether to distribute it, etc. Given that such a system sits on top of what may be critical, huge data feeds, it in many respects becomes an "owl" - a source of "truth" telling you things that, for better or worse, must be trusted. As we deploy thousands of cameras or other sensors, these systems become an important part of our working lives and our personal well-being.

The fundamental limits pertain - neither the Parrot nor the Owl system can "think" or "learn," but they can mimic both thinking and learning in situations in which we have no choice but to listen. Thinking about the "epistemology" of both "parrot" and "owl" systems is therefore highly necessary.


Fulton Wilcox
Colts Neck Solutions LLC
www.coltsnecksolutions.com

Navigating Between the Limits of Knowledge

One important fact to keep in mind is that, as we push computer-based knowledge forward, we are at risk of encountering "limits." Some limits can be overcome, while others are fundamental.

As stated by Gregory Chaitin of IBM, "... Gödel discovered incompleteness, Turing discovered uncomputability, and I discovered randomness... that some mathematical statements are true for no reason, they're true by accident. There can be no ``theory of everything,'' at least not in mathematics. Maybe in physics! "

Presumably he meant "Maybe in Physics" humorously, because given that a self-contained "logical" environment like mathematics exhibits limits, physics is in even worse shape, because it combines mathematics with real-world, often untrustworthy or incomplete experiential inputs. Modern physics demolished the neatness , knowability and seemingly infinite predictability of Newtonian "Laws."

The good news is that "determinism" is, to use a boxing term, "on the ropes" and taking a terrible beating. Not that long ago, it seemed that, collectively, the state of every particle in the universe at a given time determined the future of the universe, but the injection of limits and probabalistic ingredients along with "Chaos theory" demolishes that notion, while Einstein demolished the notion of a "given instant of time."

On the other hand, another implication to be drawn from the examples of modern mathematics and modern physics is that we can never achieve perfection nor even have full confidence in our systems at some lower limit of performance, because of these fundamental limits. Of course, in addition to these fundamental limits, we also have to handle "experiential" data, and that data brings with it further limits and contamination by error.

One major consequence is that, in dealing with the quasi-religious debates over systems design and implementation, it is important to recognize that there are no "right answers" in any absolute sense. There are what may be empirically "better" answers, but these are situational and sometimes ephemeral. Although there is a tendency among system designers to think like "Newtonians," the fundamental facts are that there is no solid "Newtonian" ground on which to stand.


Fulton Wilcox
Colts Neck Solutions LLC
www.coltsnecksolutions.com

relationship between "data" and "realworld" epistemology

The subject of this blog is epistemology as it applies to computer-centric data. Computer data epistemology is not identical to "real-world" epistemology, because the nature and features of the machine introduce certain limits and idiosyncrasies. On the other hand, it cannot be all that different and still be useful.

In the real world, we have to deal with three broad categories of "knowing:" 1) intuitive or faith-based or mytical knowledge - which in this context are more or less synonymous 2) "logical" knowledge - best typified by mathematics, and 3) experiential knowledge, perhaps beeter termed "factoid-driven" knowledge of which experimental knowledge is a disciplined subset.

One of course is not surprised that computer data is often made up from "experience," and indeed one of the reasons we employ computers is because they can be used to create, edit, store, sort and tabulate "factiods."

Computers of course were first invented as "calculators," so performing mathematical operations as well as other purely logical functions were of great importance. Note that "logical" means that the knowledge in question is internally consistent (apart from the occasional logical error), but also is disconnected from the "real world." Plane geometry supplies "proofs" regarding the "ideal" triangle, but those proofs are disconnected from the real world, because no real world triangle conforms to the essential characteristics of an "ideal" triangle.

It might seem that "intuitive" or "mstical" knowledge has no role in computer-based "knowledge." However, in many instances database structures and various other aspects of "metadata" exhibit intuitive, faith-based, or mystical characteristics. Note that, for this author, the medieval philosophic notion of "realism" and various sorts of "idealism" have relevance to today's often rather mystical debates over database design and data management.

What gets complicated is dealing with modern, large-scale use of computing, because it almost universally blends the three types of knowledge. As the 'brute force" capabilities of computer hardware, peripherals and networks increase, we are able to break out of constraints imposed by the technology - e.g., of 80 column punch cards, hugely expensive disk drives with tiny storage capacities. However, it is now that we are far less constrained by technology that we find our system designs and system operations constrained by the fundamentals of epistemology.


Fulton Wilcox
Colts Neck Solutions LLC
www.coltsnecksolutions.com

Epistemology and its impact on data management

Without "data," computer systems are merely calculators. We cannot exploit information technology (IT) without supplying relevant, computer-friendly data. Unfortunately, as computers have become more capable, we begin to encounter fundamental questions about "what we know" and how we can "know what we know."

The adage, "garbage in, garbage out" is reflective of such "epistemological" problems. Although it is long familiar, it is probably not recognized that it sometimes or even often runs "backwards" in that human insistence on "garbage out" has the peverse effect of requiring "garbage data" to be brought in.

This blog is a place to reflect on the practical implications of this topic. For more on systems epistelmology, see www.xmloptimization.com


Fulton Wilcox
Colts Neck Solutions LLC
www.coltsnecksolutions.com