AI - supporting decision making

Machine learning is built on algorithms to learn and provide results to end user (Chavan, Somvanshi, Tambade & Shinde, 2016). It is considered as a subfield of artificial intelligence. Complex mathematical, statistical and computing models are developed and tested to train a machine which can later help in generating inference for new unknown data. Some of the everyday examples include search engines, spam filtering, speech recognition etc. On a higher level, machine learning is classified into three categories:

1. Supervised learning – training data is supplied to algorithm and then validated against known outcomes. Once model is tuned, new and unknown data is supplied for results

2. Unsupervised learning – some input data is provided which is used by algorithms based on statistical models for classification

3. Reinforcement learning – involve online performance optimization by balancing between exploration of unchartered territory and exploitation of current knowledge

In the era of big data, artificial intelligence algorithms are widely used to extract intelligence and make decisions (Earley, 2017)). However care must be taken as there are data integrity issues with missing or incorrect values which can sway the results in wrong direction. Data scientists tend to spend quite a time in just cleansing the data to avoid inaccurate results which require significant amount of knowledge engineering.

One of the example of AI decision which was not correct happens in the 2014 shooting event in Sydney where numerous calls to Uber increased the rates due to supply and demand with no consideration of circumstances (Ruoco, 2017). From the aspect of AI in class room environment, technology addiction is one problem but also lack of personal connection can lead to some degradation in overall quality (LiveTiles, 2017). Reddy (2017) also highlighted that it is complex to let AI evolve with the passage of time and learn from experience to enhance the quality of decisions. Similarly highly sensitive and emotional intellect of humans can bring creativity which is not the characteristic of AI for now.

In summary, AI should support decision making and be used as a tool for validation or second opinion. AI can also perform difficult and repetitive tasks quicker on huge data sets for predictive modeling thus reducing the time to support final decision making.

Reference:

Earley, S. (2017). The problem with AI. IEEE Computer Society - IT Professional, 19 (4), 63-67

Chavan, P., Somvanshi, M., Tambade, S., & Shinde, S. (2016). A review of machine learning techniques using decision tree and support vector machine. 2016 International Conference on computing communication control and automation.

Ruoco, K. (2017). Artificial Intelligence – the advantages and disadvantages. Retrieved from https://www.arrkgroup.com/thought-leadership/artificial-intelligence-the-advantages-and-disadvantages/

LiveTiles (2017). 15 pros and 6 cons of artificial intelligence in the classroom. Retrieved from https://www.livetiles.nyc/blog/pros-cons-artificial-intelligence-classroom/

Reddy, K. (2017). Artificial Intelligence – advantages and disadvantages. Retrieved from https://content.wisestep.com/advantages-disadvantages-artificial-intelligence/