Dasein's Edge on its Description
Hans H. Diebner
Institute of Basic Research, Center for Art and Media, Karlsruhe, Germany
Manuscript as of November 21, 2004
The following considerations have been stimulated by observing a change of opinion of physicists on subjective probabilities as they have been widespreadedly used in medicine for a long time. This is interpreted as a latent commitment to hermeneutics. It is, however, at the same time a venturous claim to „algorithmize hermeneutics“. This mission has an alarming potential to outshine the advantages of artificial intelligent design. We are exposed to a sneaky enslavement by „prejudice confirming“ algorithms based on tracked information on our previous decisions. On the other hand, the algorithms' results are approximations to our decision-making only. Dasein has an edge on its description. A competent handling based on knowledge, however, is necessary.
Traditionally, the natural sciences do not count medicine to one of their disciplines. This is, in a nutshell, due to the dependency of diagnostics on subjective probabilities – via the „Bayesean inference principle“. Statistics is devided into two schools of thoughts, the Bayesians and the frequentists. The latter exhibits a certain similarity with structuralism in philosophy whereas the Bayesean school can be compared with hermeneutics1. The diagnostics procedure starts with an a priori probability as a degree of believe in a hypothesis which is mapped to an a posteriori probability. This mapping results from a necessarily limited experiment, observation or test. In order to estimate the efficency of the observation one tests the test to derive measures called „sensitivity“ and „specificity“. This „meta test“ depends on an absolutely reliable comparative test called „golden standard“. Needless to say, the golden standard goes back to an earlier experience – an iterated application of Bayes' inference. One sees the emerging aporia. Plato coined the term anamnesis for the soul's innate capability to squirm from an aporia (a logical perplexity). In medicine, anamnesis, the thoroughly ascertainment of the previous history, is the modern variant of Hermes' message „out of the blue“ or to start from somewhere in the middle.
A considerable number of cognitive scientists believe in a „Bayesean homunculus“ iteratively acting in our brain to muddling through the deciding processes we are continuously exposed (Gigerenzer, 2000). We decide on the basis of limited sensitivity and specificity, in short: on „priors and prejudices“. In the specific case of assessing an x-ray image of a patient the physician becomes identical with the diagnostics test. One would run into a trapped, self-referential problem if the „test“ result were scrutinized in the same way as normally done with an external measurement equipment where a prediction value is acquired to estimate the validity of the result. No wonder that algorithms of the iterated Bayesean type are called „bootstrapping methods“ and associated with hermeneutics (Mallery / Hurwitz / Duffy, 1987).
Astonishingly, in modern physics and cybernetics the Bayesean inference principle has been recognized as a powerful algorithm for adaptive processes which is why artificial neural networks are endowed with „Bayesian learning“ (Cozzio-Büeler, 1995). Suddenly the medical way of gaining knowledge not only became accepted but moved to the center of cyberneticists' phantasies to triumph with artificial intelligence over mankind (Kurzweil, 2000). Recent developments in surveillance software, organizational efficiency programs, evalutation tools for solvency, and others remind us to take this evolution very serious (Silberman, 2000). Social disposition evaluated with Bayes' algorithm adds to the genetic disposition. Of course, neither the genes nor the accumulated results from our social behavior fully determine us. It is an astonishingly good approximation, though, which leads to a dangerous unawareness of the techniques at-hand since we always find our prejudices confirmed. The little helpers in the background, frequently called bots – the externalized Bayesean homunculi – like the Microsoft office assistent or spam filters absolve us to think ourselves. It is a sneaky enslavement.
Physicists and mathematicians, the former critics of the physician's way of gaining knowledge, act up to have the better recipe for diagnostics. Radiological and/or tomographical images are analyzed with the aid of artificial intelligent systems, under medical consultancy in a first approach. In the course of time it is intended to gradually substitute the physician and leave the diagnostics task up to the „objective“ algorithm. In a 1971 textbook for physicists an interesting side remark has been made which goes as follows. In general, for the Bayesian the probabilities are the degree of belief in a hypothesis. The anti-Bayesean objection is that all scientists will have different degrees of belief, and so the conclusion will be subjective. The Bayesian's defence is that the a priori probabilities should contain all hypotheses and all previous knowledge, and that if all scientists would pool their previous knowledge, they should be able to agree on a distribution for the a priori probabilities (Eadie / Dryard / James / Roos / Sadoulet, 1971). In medicine this organizational idea evolved into a widely propagated programme called „evidence based medicine“ (Sackett / Straus / Richardson / Rosenberg / Haynes, 2000). Of course, it has not been foreseen 15 years ago when this programme was created that it will provide also the bases for organizational sciences and knowledge management with the attempt to measure the unmeasurable. The keywords are „competence“, „expertise“, „soft skills“ and so forth. Research seems to spin round since competence is revealed in acting which detracts from quantification.
From socio-biological studies R. Axelrod provided a hint for the limitation of the Bayesean approach to cognition (Axelrod, 2000). He organized a tournament of algorithms which pairwise had to compete and to prove themselves in a fundamental decision making process called „prisoner's dilemma“. Hereby, the decisions have to be made under „bounded rationality“, that is with restricted knowledge on the situation. The efficacy of the subsequent decision depends upon the subsequent decision of the antagonistic algorithm and vice versa. In other words, the efficacy stands and falls with a good estimation of what the other algorithm will do. One of the remarkable outcomes was that in a first round the Bayesean algorithm did a pretty good job. However, in a second round after the creators of the algorithms could analyze the first round, a tricky new algorithm estimated the outcome of the other algorithms including the Bayesean to find the most efficient counter strategy. It is exactly this creation of new hypotheses that definitely cannot be captured even by so called self-modifying algorithms so far. The mathematical modelling of human behavior is always in retrospect. It is ergodic. Time can be transformed away. In a way this modelling deals with equipment not with processes. „Dasein“ has an edge on its description.
System theory is nowadays aware of its limitations and speaks in a fuzzy term of „emergence“ when attempting to describe the new (Crutchfield, 1994). The paradoxes of the new has been discussed much earlier in philosophy by Whitehead and Heidegger, for example (Bradley, 2004). To approach the problem via system theory provides a new perspective on this philosophical settings and probably even leads to new insights. Although „the hermeneutics approach stands where artificial intelligence fails“ (Fonseca / Martin, 2005), philosophy and the sciences can benefit from each other. Upshur reversed the concept of „evidence based medicine“ to an „evidence based hermeneutics“ (Upshur, 1999). We glue this concepts together to an „operational hermeneutics“ (Diebner, 2003) that, firstly, regards natural sciences as a hermeneutic discipline and, secondly, regards sciences as the optimal tool for – but only for – a retrospective description, that is for ergodic, a-historical aspects of Dasein and, thirdly and most important, regards technics as a source for the creation of new hypotheses. Note, that technics (artificial intelligent systems, algorithms, ...) cannot create hypotheses by themselves. „Operational“ stands for a discourse that comprises cybernetic design. Operational hermeneutics continuously switches between both half spaces – the ontic and the epistemic one – of the semantic generating surface – the interface (Diebner / Druckrey / Weibel, 2001). It bridges the cultures.
Axelrod, Robert (Ed.) (2000): Die Evolution der Kooperation. München: Oldenbourg.
Bradley, J. (1991): Whitehead, Heidegger, and the Paradoxes of the New. Process Studies 20, 127-150.
Cozzio-Büeler, R.A. (1995): The Design of Neural Networks Using a priori Knowledge, Diss. No. 10991, ETH Zürich.
Crutchfield, J.P. (1994): Is Anything Ever New? Considering Emergence. P 479 – 497 in: G. Cowan / D. Pines / D. Melzner, Complexity: Metaphors, Models, and Reality. Reading, MA: Addison-Wesley.
Diebner, Hans H. / Druckrey, Timothey / Weibel, Peter (Eds.) (2001): Sciences of the Interface. Tübingen: Genista.
Diebner, Hans H. (2003): Operational Hermeneutics and Communication. P 30 – 57 in: Hans H. Diebner / Lehan Ramsay (Eds.), Hierarchies of Communication. Karlsruhe: Zentrum für Kunst und Medientechnologie.
Eadie, W.T. / Dryard, D. / James, F.E. / Roos, M. / Sadoulet, B. (1971): Statistical Methods in Experimental Physics. Netherlands: North Holland.
Fonseca, F.T. / Martin, J.E. (2005): Toward an Alternative Notion of Information Systems Ontologies: Information Engineering as a Hermeneutic Enterprise. Journal of The American Society for Information Science and Technology 56(1), 46 – 57.
Gigerenzer, Gerd (2000): Adaptive Thinking: Rationality in the Real World. New York, Oxford: Oxford University Press.
Kurzweil, Ray (2000): The Age of Spiritual Machines: When Computers Exceed Human Intelligence. East Rutherford, NJ: Penguin Books
Mallery, J.C. / Hurwitz, R. / Duffy, G. (1987): Hermeneutics: From Textual Explication to Computer Understanding? P. 596 – 611. In: S.C. Shapiro (Ed.), The Encyclopedia of Artificial Intelligence (Vol. 1, 2nd edition). New York: John Wiley & Sons.
Sackett, D.L. / Straus S.E. / Richardson W.S. / Rosenberg W. / Haynes R.B. (2000): Evidence-Based Medicine How to Practice and Teach EBM (2nd edition). Edinburgh: Churchill Livingstone.
Silberman, Steve (2000): The Quest for Meaning. Wired 8.02, 172 – 179.
Upshur, R.E.G. (1999): Priors and Prejudice. Theoretical Medicine and Bioethics 20, 319 – 327.
1 Medicine is approved as a hermeneutic discipline. See e.g. Upshur (1999) and Diebner (2003).