Appendix B — Relation to AI
The four areas of data science defined here are surprisingly analogous to the four approaches to artificial intelligence defined by Russel and Norvig in their classic textbook on the subject (Russell and Norvig 1995: 5). Their four part model was generated by combining the axes \(thinking \leftrightarrow acting\) with \(human \leftrightarrow rational\) as follows:
| behavior | mode | systems |
|---|---|---|
| thinking | humanly | cognitive models, ontologies |
| thinking | rationally | logic, laws of thought |
| acting | humanly | Turing test, situated action |
| acting | rationally | agents, bots |
Now, it is easy to see how the following analogies make sense:
\[ abstract : concrete :: thinking : action \]
and
\[ human : rational :: human : machine \]
Or, put in terms of a set of identities:
| AI | DS |
|---|---|
| abstract | thinking |
| concrete | action |
| human | human |
| rational | machine |
So it makes sense to compare the above table with this one for data science:
| level | focus | area | topic |
|---|---|---|---|
| abstract | human | design | ontologies, data models, visualizations |
| abstract | machine | analytics | math, logic, algorithms |
| concrete | human | value | ethics, research questions, value propositions |
| concrete | machine | systems | hardware, software, security |
By comparing last columns of each table we can see that the 4+1 model of data science and Russell and Norvig’s model of artificial intelligence share the same general space. The difference is that the former defines kinds of acquired knowledge, whereas the latter concerns kinds of built systems.
References