Cognitive Biases – Invisible Impact on your Data Science Project
We are living in a world with plenty of intelligent minds. At workplace also, there are many intellectual minds, experienced minds working together with each other on a single project/goal to achieve best for the team/company. As a Data Scientist, we will be an integral part of that team and the ultimate result of our work, our statistical or ML model on the business use-case will be decided by all these higher officials. Now, we know that the officials will take the decision best for the business, but we must also consider that the officials are simple human, and thus their decisions and opinions are sometimes biased, reason can be anything or may be unknown sometimes. And being biased is normal for a human being, but as a Statistician, Mathematician or Data Scientist, our work is to take decisions based on facts and figured and not being biased by something. This is an invisible influential concern for us, and we will be discussing about that in this article.
Our mind’s bias is called “cognitive bias,” a systematic error in thinking that occurs when people favor their personal experience into a decision without properly putting it on the merit scale.
There are countless biases that are difficult to avoid, but being aware of them allows the data scientist to take them into consideration.
Common Cognitive and Motivational Biases
Anchoring Bias
The first common cognitive bias is the "Anchoring Bias", which occurs when the estimation of a numerical value is based on an initial value(anchor), which is then insufficiently adjusted to provide the final answer.
In simple words, the "Anchoring Bias" occurs when the decision is highly or majorly influenced by the first piece of information, somewhat neglecting other might be more important features of the problem or the situation.
For example, imagine that you went for shopping and you saw the "Sale" board, and immediately you are attracted towards the items at a discounted price. But, actually the store owner played with you "Anchoring Bias" influenced decisions. They hiked the prices by a certain percentage and then added the discount tag, showing you that the price is discounted but actually that price is the real selling price for that store.
Similar, situation can happen with you during the decision making rounds at your work place, and thus to get the real outcome and to make the required results we must avoid the "Anchoring Bias" during decision making.
De-biasing Techniques
Taking the decision on the basis of the first information available is never recommended. We can take care of certain things to avoid the "Anchoring Bias"
Affect Influenced
This bias occurs when there is an emotional predisposition for, or against, a specific outcome or option that taints judgements.
For example, imagine a situation where you are presenting your recommendations to the stakeholders, and you boss knows that you are intelligent and have performed very well in previous projects and thus he immediately decides to go forward with your recommendations, whereas the CEO do not know much about your previous achievements and is tensed to take risk and thus decided to listen some other officials opinions and decide after that. Here, you boss has taken the decision based on "Affect influenced" bias, unlike the CEO.
De-biasing Technique
Ambiguity Aversion
This bias occurs when people tend to prefer gambles with explicitly stated probabilities over gambles with diffuse or unspecified probabilities. Which mean, people tend to take calculated/known risk over ambiguity in taking that risk or taking the unknown risk.
De-biasing Technique
Equalizing Bias
This bias occurs when decision makers allocate similar weights to all the objectives. It is important to allocated desired weights to different objectives of a project to yield the desired result and we can avoid this bias by taking certain steps during our work.
De-biasing Technique
Recommended by LinkedIn
Confirmation
This bias occurs when there is a desire to confirm one's belief, leading to unconscious selectivity in the acquisition and use of evidence. The person tends to confirm what he believes, unconsciously neglecting the facts.
De-biasing Technique
Base Rate Fallacy
This bias occurs when people tend to ignore base rates when making probability judgements and rely instead on specific individuating information. The example can be seen in the image above.
De-biasing Technique
Desirability of Options
This bias leads to over- or under-estimating probabilities, consequences in a direction that favors a desired alternative. The de-biasing technique for this one can be seen below.
De-biasing Technique
Insensitivity to Sample Size
This bias occurs when people tend to ignore sample size and consider extremes equally likely in small and large samples. This de-basing technique for this one can be seen below.
De-biasing Technique
CONCLUSION
These were some main cognitive biases one must take into consideration while working on his/her Data Science project or any other domain also. The biases are mostly invisible when they occur but the effects of these biases are significant on the end result. Thus, we must always consider be-biasing these to get the accurate result of out work.
References