Artificial Intelligence (AI), Data Science, and Analytics

There is a lot of confusion about the definition of Artificial Intelligence (AI), Data Science, and Analytics, and it is particularly harmful for students and early-career people who are thinking about specializing in these areas. This article is to dispel that confusion.

Artificial Intelligence (AI) is characterized as “intelligent software agents” in the authoritative book on the subject, “Artificial Intelligence: A Modern Approach” (Russell, 2010). Advances in machine learning, computer vision, and natural language processing create a buzz for AI. This buzz sometimes overshadows other AI algorithms, such as search, stochastic games, etc.

Data Science is about organizing and analyzing massive amounts of data. It is tightly linked to machine learning (ML) tools. These tools are also called AI. The overlap with AI is often resolved as “AI makes the tools; Data Science uses the tools”. If so, Data Scientists are the mechanics who train and maintain Machine Learning (ML) algorithms, not the engineers who design and build the system. This is at odds with the aspirations of those Data Scientists who aim to “extract value from data” (Irizarry, 2020), and forces a redefinition of Data Science towards becoming an umbrella term highly overlapped with Analytics.

Analytics is an umbrella term for all that it takes to “transform data into insight for making better decisions” (Saxena, 2020), and includes the tasks to:

1. Create analyses using statistics, operations research, and decision analysis. Present results in reports, scorecards, and dashboards.
2. Build and run data supply chains. Manage data and metadata, assure data processing flows, and data quality.
3. Design and implement user interfaces, dashboards, reports, etc.
4. Code and maintain data processing, system interfaces, and storage.
5. Set up and run the systems infrastructure of servers and networks.

Analytics systems generally get data from transaction systems. Transaction systems are the primary originators and sources of data. There are many kinds of transaction systems, such as Enterprise Resource Planning (ERP), Customer Relationship Management (CRM), Supply Chain Management (SCM), etc. Other data sources, such as Social Media (Facebook, Twitter, etc.), the Internet of Things (IoT), etc., may also be put into this “transaction systems” category for the purposes of this discussion. The difference between analytics systems and transaction systems rests on the differing purposes: one to analyze, the other to do a transaction or task.

Analytics systems and transaction systems need not use AI, they can and often do use non-intelligent software. We can, therefore, categorize all systems as intelligent or dumb depending on whether they use AI or not.

• Dumb systems. These systems are used for data input and storage, with basic “smarts” in data processing such as edit-checks, filtering, grouping, summing, etc. For complex systems such as ERP and CRM, the intelligence is often in the minds of the experts who configure and customize the systems based on their evaluation of business needs, after which users can use the dumb system along the exact pathways set by the experts.
• Intelligent systems. These systems embed the “mind of the analyst” in the software, enabling the system to achieve its goals (e.g., to transcribe voice to text, or to build the optimal staff roster) for users who do not have any expertise in the algorithms being used.

Systems in a 2x2 matrix: Transaction/Analytics vs. Dumb/Intelligent Systems

Now we can define AI, Data Science, and Analytics in the context of this 2x2 matrix representing the universe of systems.

• AI deals with designing, building, and running intelligent systems — both analytics and transaction systems.
• Analytics deals with designing, building, and running analytics systems — both intelligent and dumb systems
• Data Science deals with tuning intelligent systems in a “Machine Learning Engineer” role but may (and often does) aspire to AI or Analytics roles.

References

Russell, S. & Norvig, P. (2010). Artificial Intelligence: A Modern Approach. Prentice Hall. ISBN 0134610997.

Saxena, R. & Srinivasan, A. (2013). Business Analytics. Springer. ISBN 9781461460800.

Saxena, R. & Gupta, R. (2020). The Analytics Asset. In Global Business Leadership Development for the Fourth Industrial Revolution, Smith P. & Cockburn T. (Editors). IGI Global. ISBN 9781799848615.

Irizarry, R. A. (2020). The Role of Academia in Data Science Education. Harvard Data Science Review. DOI 10.1162/99608f92.dd363929.