Please read below and reply to the discussion with your opinion. No citation is needed.
Size or accuracy of data are not the only issues that decision makers and data scientists should look out for when conducting data analyses. It is proven that peoples own misjudgment, expectations, beliefs, confidence levels, and pressure to conform to bosses wishes are issues to look out for when assessing the integrity and ethical process of the analysis. When handling data it is important to stay neutral and avoid prejudice in order to achieve honest, accurate results. In chapter 15, the author discussed some of the issues (or cognitive traps as he calls it) that might have a negative impact in data analysis.
Issues
The confirmation trap Sometimes we start analyzing data that aligns with our beliefs and ignore data that contradicts them. This is called the confirmation trap and is one of the biggest issues when data driven results are needed. This issue wrongfully narrows down or limits the number of variables that may influence outcomes. Sometimes it may arise from subordinates desire to please bosses or higher executives since employees know what they expected or wanted results are. This practice avoids basing the analysis on empirical evidence and brings bias as a consequence.
The overconfidence trap When senior decision makers have a history of promotions based on past successes, they might develop overconfidence. Although confidence is wanted in order to achieve higher results, overconfidence may have a different effect. It can lead decision makers to only value their way of doing things and dismissing other methods or approaches. If this issue worsens, it may also result on the decision maker underestimating others work, experience, skills, results, and only valuing their own.
The overfitting trap This issue arises when relationships between two variables are highly pursued or exploited to overfit one another. This is an issue since the analysis is so focuses on the variables being observed that it might dismiss other variables or correlations. In the worst scenario, overfitting might cause you to miss or overlook important data or underlying relationships that are relevant to the analysis.
Lessons learned
Heuristics are methods, strategies or procedures that are developed from prior experience in dealing with a similar problem. Unlike machines, people not always stay neutral when working with data and might conduct their entire data gathering/analysis process based on undesired, opinionated ideas. This takes away the honest procedural way of conducting data research, which is starting with data and reaching to a result. Heuristics might cause data scientists to work backwards by starting with a result in mind and trying to do everything to make it fit to the available data.
Pressure is a strong motive for employees to use heuristic methods in order to satisfy employers expectations. Employees might feel the need to content their superiors by affirming their ideas on results or outcomes. Sometimes, handling data in a dishonest data corresponds to employees desire to earn a higher salary or bonuses if the results conform to a plan or strategy drafted by the company.
Managers should take into account that data analyses might contain biases and should set a guideline in order to avoid them. Although some of these issues and bias inadvertently occur when conducting data analysis, management should also penalize intentional miss-handling of data in order to fit a personal purpose or need. Management should promote data-driven decision making and encourage neutralism when conducting research.
Best practices
Confirmation can be embraced by specifying in advance the analytical approaches that are going to be employed when conducting analyses. Also, actively looking for findings that disprove ones beliefs might be a good way to avoid the temptation of agreeing with personal prejudices. The author suggests to not overlook or dismiss that fall below your threshold and to assign multiple teams to analyze the same data independently.
To avoid falling into the overconfidence trap, the author offers some procedural tips including the description of a perfect experiment. Before making a decision, a pre-mortem should be conducted in order to benefit from various perspectives and identifying potential flaws. Keeping track of predictions and comparing it to actual results is another approach that can help catch biases and prevent them from an early stage.
Preventing the overfitting pitfall, one might divide data into two sets (a training set and a validation set), which are helpful in making predictions and raising flags. Checking the hypothesis before you jump into analyses is another way to keep the trail of what it is we are working with. Sometimes, its also important to construct alternative narratives in order to analyze whether there are other outcomes that arise from the same data.
Relation with Topics Covered in Class
These analyses relate to the topics covered in class since we tend to use data to our own benefit sometimes instead of following a neutral approach when conducting research. In the team projects we had to use data that were given to us and use it to arrive to results. If we were to include our own judgmental opinions of how data fit to a certain idea, we would be including bias into our analyses and not reporting a genuine result.
Alignment of Concepts Described in the Chapter and Concepts Reviewed in Class
These concepts are similar to integrity. For example, in the case of overfitting, employees sometimes cheat in data analyses in order to arrive to a desired result instead of conducting an honest research. This is highly unethical and dishonest.
Chapter 22
Chapter 22 is a really brief, short chapter that describes how decisions dont start with data. When people buy products or make decisions as to which company theyll contract with, they need numbers and information in order to make a good choice, but there are other factors that also play a role. How well data is presented and how strongly this data is aligned with a story are key factors when appealing to an audience and make them interested in your product.
Issues
Sometimes companies have data availability, accurate methods, good techniques, and powerful data analytics. They show strong results backed up by data; however, they never seem to progress or achieve goals as they have strategized. Whats the issue then? It may have to do with the way they are portraying that data to outsiders of the company. People that work for the company may know the opportunity for growth and success in the company, but outsiders need a little bit more than numbers to trust in your company. The issues are the image presentation and the lack of a story telling approach in order to captivate investors, creditors, potential partners, or consumers.
Lesson Learned
Although data is helpful at supporting material, the author makes a strong point on how numbers dont speak for themselves. Effective persuasion is needed when appealing to a group of people. Numbers are there as a way to support and back up your story or information; however, one should never just make a case based off numbers only without including emotional power.
Best Practice
Data provides insights and offers and deliver deep understanding of relationships, outcomes, and trends. However, data does not convince or persuade others. People do. Therefore, it is important that people presenting data have strong communication skills, persuasion skills, and the ability to keep the audience interested in the story.
Relation with Topics Covered in Class
We have seen in past discussions how data may influence marketing. This applies when advertising. Even though people rely on numbers and facts, people dont really buy products just because there is data backing up the information. People buy products when they feel it has a deeper meaning, connection, or emotional relation attached to it. We see how commercials use communication and emotional appeals in order to persuade potential buyers, attract a greater costumer base, and generate higher profits.
Alignment of Concepts Described in the Chapter and Concepts Reviewed in Class
In our team cases, we not only had to provide numbers, outlines, and excel reports, but we also had to speak from a sympathetic standpoint in order to attract the attention of PLE. We used recommendations as a way to persuade them in one direction or another, and kind of tell them a story of how those numbers proved that the company could do so much better in the future if some steps were taken.
