Select Page

Algorithms as part of artificial intelligence is a much talked subject during the last years, although little explanation is given about the functioning and effects of the procedures. Basically, it’s a more or less complicated mathematical method to find fast solutions for complex problems.

In 1975 professor John Holland (Michigan, 1929-2015) began to use the concept of biological evolution to create a simulation model. First, this model was a very good problem solving procedure, being able to detect almost optimal solutions even in complex environments. Second, the mechanisms working in Genetic Algorithms (GA) seemed to be common mechanisms of evolutionary processes in general. They are the mechanisms explaining the origin of novelty and why systems undergoing evolution are able to adapt so well to their environment. Both aspects seemed to qualify them as a means to describe economic evolution or learning and adaptation of economic agents.

Nowadays algorithms are made for and used in almost any sector of the industry, health care, and in the economic world. In the area of brain research, for example, algorithms have become indispensable devices. The last 25 years science has developed rapidly in the fields of genetics and neurology. The effects of specific genes on our body are becoming clear. We can retrace certain bodily functions and pathogens to their genetic origins. Magnetic resonance imaging (MRI) scanners produce detailed images of any part of the body. They can be used to help diagnose conditions and plan treatments. In brain scans they detect brain tumors, traumatic brain injury, developmental anomalies, multiple sclerosis, stroke, dementia, infection, and some causes of chronic headache. They are also used in research to brain activity as a result of certain inputs. This neuroimaging technique allows us to detect the specific areas of the brain which are involved in a task, a process, or an emotion.

Also, in the judiciary algorithms play a decisive and maybe questionable role. Currently, courts and corrections departments around the US use algorithms to determine a defendant’s “risk”, which ranges from the probability that an individual will commit another crime to the likelihood a defendant will appear for his or her court date. These algorithmic outputs inform decisions about bail, sentencing, and parole. Each tool aspires to improve on the accuracy of human decision-making that allows for a better allocation of finite resources.

In financial markets and economics in general algorithms determine the course of action. Genetic algorithms are used in firms to learn to estimate changes in sale and prices over a certain period. When using GA in an economic context, the operators of the algorithm are interpreted as steps of a process of adaptive learning. A string is an idea or a market strategy of an agent. Selection and the related fitness function depend on the agents’ success in their economic environment, and on the agents’ personal interpretation of the reason for this success. New ideas are developed by recombining or copying successful old ones (crossing over), including some mistakes or experiments (mutations).

In the economy the same selection process takes place, although in a mixture of intentional internal selection and external selection by the market. The market selects products for their selling success, reflected in prices, sold quantities or market shares. This is the basis for the intentional selection by entrepreneurs or inventors, when designing new products.

 Health care. Algorithms substituting for medical judgment? Vinod Khosla, the luminary venture capitalist and the co-founder of Sun Microsystems, shook the technology and the medical communities with his highly talked about article, “Do We Need Doctors or Algorithms?” In the article Khosla argued that given the level of service that we seek and eventually receive from 80 percent of physicians, we might be better off receiving that care from a computer with sophisticated algorithms. Khosla fondly named that system “Dr. Algorithm” or “Dr. A,” for short. Later in his follow-up talks he ignited the debate further by saying that 80 percent of physicians in the United States can be replaced with machines, and that day is not very far away.

The medical community responded with the argument that healthcare is not about technology – it is about the intersection of technology, science and human emotions, along with the therapeutic touches and listening abilities of a doctor. Decision support software, therefore, seeks confirmation from the clinicians. Machines also don’t have access to the huge data in healthcare that is needed to generate the desired precision in diagnosis. Genomic data is sporadic, and the majority of the clinical encounter data is still not digitized. Further complicating matters, when electronic data is available, the absence of data liquidity and interoperability within and among healthcare organizations makes it harder to get a holistic view of any patient.

The key here is that physicians have to let the machines learn from their decisions or mistakes, and as IBM is finding out, that is non-trivial. How do you scale when every project is custom built, takes a long time to complete and yet you are at the mercy of the physicians who fear that they are training their replacement? Moreover, even when personal medical data is available patients are concerned that seamless data flow among healthcare stakeholders will destroy their privacy and make them more vulnerable to insurance payers and employers. Not an easy problem – is it?[1]

The objections from the medical field are understandable and to the point. Two advantages of algorithm driven healthcare might be that results, diagnostics and type of treatment, come to light much faster, and that it is probably less expensive than the personal doctor’s approach without A.I.

What might be said in general about the application of algorithms is that most are created by private companies, who don’t divulge the content. This lack of transparency is understandable because their products are practically not patentable (especially in Europe). Apart from that the user (and the Patent Office, if applicable) must have sufficient mathematical knowledge to even begin to comprehend the software. A more sentimental objection would be that we gradually degrade ourselves to merely enforcers of non-thinking machines.

[1] This paragraph is partly cited from an article from Rajib Ghosh in Analytics (April 2014).