When discussing the use of algorithms, the issue of durability or portability has to be considered. For example, a stock trading algorithm might be used in a missile guidance system. The algorithm would have to operate on an abstract kinetic level rather than for a specific application. I have written in the past about using the same algorithm to study stocks, earthquakes, hurricanes, electro-cardiograms, and attempts at evasion - using my mouse in a game environment. Wouldn't an abstraction be limited first by an incomplete understanding of phenomena and second by the problems posed by inferential modeling? I am not a seismologist or geologist. I consider my understanding of earthquakes insubstantial. By logical extension, I cannot provide a coherent explanation to extract an algorithmic model from earthquakes or to justify as analog an application from a foreign source. I also doubt that many people would read the research that I do if I lack an appropriate background in the particular field of study. I consider this a unique problem that likely plagues some data scientists given the focus on the data rather than specific research discipline. We can laugh it off when Google shows its inability to predict the spread of disease; this area of study isn't really the company's area of specialization. Nonetheless, setting aside jealousy, I still find myself thinking that any tangible level of success would have been amazing. It would have been good for the community.
Or would it? I think there is still great debate over who belongs to the community and what people within it do. Inferential modeling in particular doesn't really belong to any specific discipline. Development by inference is a great skill. For example, associating search engine searches for the flu with actual cases of the flu is the sort of inference that might be reasonably made - by a 12-year-old. It makes sense, but it doesn't take much thought. Associating searches for terrorists - or any so-called "chatter off the net" - with terrorism likewise makes sense. This is the "If I build a landing strip, the enemy will land there" sort of reasoning that really seems quaintly childlike. If that is all it takes to portray a picture, terrorists only have to spam the internet with "talk of terrorism." This in no way reduces my desire to effectively deal with the problem. I simply recognize the limits of using algorithmic metrics based on inferential models. Sometimes, it isn't a matter of lack of sophistication in the algorithm but rather wrong-headedness in the inference. I have mentioned in the past that I consider the "science" part of "data science" to be the testing of metrics against outcomes. I recognize that I might be alone in this perspective. But a real data scientist reports the extent to which the flu can be predicted based on specific search parameters; a quasi-intellect simply asserts the metrics and expects reality to adhere to the model.
Understandably, algorithmic discourse emerges in relation to abundant amounts of data. Due to the prevalence of mathematics in society, we have a means of "plugging in" different algorithms to streams of data. However, I find it useful to apply "algorithmic reasoning" even on limited amounts of data. Certain applications exist in relation to business: small scale experiences might be asserted on larger scales (the so-called pilot project); foreign experiences are used to guide local; hypothetical case studies are applied to real situations. The successful application of an algorithm depends on the availability of analogs on which to test the level of congruence or symmetry. While I realize that the community might not recognize how algorithmic reasoning on a small scale might contribute to data science, nonetheless this type of reasoning process can bring about the development of algorithms. Having algorithms that can be applied to real-world problems having real-world scale is important; it prevents foolhardy assertions that seem entirely dissociated with reality. This is not to say that being alienated from reality is necessarily all that bad in terms of scenery; it merely has an adverse impact on reputation and public confidence.
I mentioned in a previous blog that I recently took up Tai Chi. Tai Chi has been around for several hundred years. It represents an entry point to Qigong - around for several thousand years. Qigong has been used in the treatment of disease. If all of this seems unbelievable, replace the words "Tai Chi" and "Qigong" with "exercise." Then suddenly it's fine, right. A search engine might not be designed to recognize the differences and similarities. Similarly, a person exposed only to Western culture might fail to recognize worthwhile analogs. Computers have only been around for little while; it can be tempting to dismiss everything that proceeded. As a result of my graduate research into workplace stress, I find all sorts of analogs between the following: Tai Chi; Qigong; faith healing; glymphatic system; lymphatic system; sympathetic system; parasympathetic system; serotonin; melatonin; pineal gland; proprioception; dysautonomia; obesity; and of course workplace stress. I became aware of some of these analogs due to the phenomenological significance of writings from ancient Qigong practitioners. There are also completely Western expressions that draw my attention: "I had chills running up and down my spine." Is this an expression only; or did a high-stress environment invoke the release of substances in anticipation of a flight response? "I felt the Holy Spirit enter into me, and I knew I was cured." Well, if we dismiss the potential analogs, there are no connections. On the willingness to investigate analogs, it is possible to troll the net either to develop or test inferential models.
In short, there are different ways of coming up with ideas (inferential models) to associate with search parameters (application of analogs). I spent a moment sharing an area that I am extremely interested in - although I appreciate I might be the only person. Nonetheless, I think that the general idea is important in relation to data science. As I mentioned earlier - and which I didn't quite cover in this blog - there is the question of legitimacy due to absence of professional background. People wouldn't mind me writing about stress particularly in relation to social conditions; after all, I received a degree doing just this. But the validity of deploying technology in a systemic manner over massive amounts of data remains to be demonstrated. It is difficult to develop a discipline that is so easily delegitimized due to lack of domain expertise. An algorithm would be quickly proven ineffective and its researchers tossed into the fires of Gehenna without much fanfare. What we cannot gain through numerical advantage we should probably attempt to achieve by providing insights. Sometimes, an algorithm is less useful for predicting the future than establishing likely leads and connections for scientists to confirm. With search engines and online documentation becoming so important to our social and technological development, there is an enormous untapped frontier ahead of us to develop systems of insight and social awareness.