Critical thinking and decision-making play a core role in dealing with a project, as they can make your project successful or unsuccessful. So, You must learn both of these integral parts of data science if you want to get amazing and efficient results.
Several mental models in data science will help you polish your decision-making and problem-solving, and we will guide you about everything in this article.
It’s time to get your hands on these skills and explore a new world to deal with your problems like a pro.
What are Mental Models in Data Science?
Mental models are our understanding of the outer world. It’s about how we understand the world and interact with it.
In Data Science, you deal with a bundle of problems and need to make specific decisions and communicate them to others (Your Team). Mental models help in optimizing your decision-making and problem-solving.
Why Use Mental Modes in Data Science?
Below are some of the reasons you should use Mental Models:
- You can think better & efficient
- You can make better decisions
- Easy Problem Solving
- You can simplify your problems
- Work Quality Increases
- Gives you clarity
What are the Drawbacks of Mental Models in Data Science?
There are a few cons of using Mental models in data science as well: (But when we compare pros with cons, Mental models always win)
- Mental models are not always perfect
- Not all problems can be solved using Mental models
- You have to spend many hours thinking
- If wrong mental models are used, your project may become more problematic
- Mental models only work in certain situations
What are the Decisions that a Data Scientist makes?
Sometimes data scientists have to manage business and technicalities simultaneously, and it’s not easy to make decisions for most data scientists. In this situation, data scientists can get benefit from mental modeling.
Some of the Decisions are:
- Goal Setting
- Work Scope
- Responsibility, and so on…
Common Problems that a Data Scientist Face While Making Decision
A data scientist may encounter several issues while making a decision. We have enlisted a few issues a data scientist faces:
- Handling a team
- Collecting the right data
- Not enough information
- Pressure from Stakeholders
- Unrealistic requirements
- Unusable data
If you’re also facing these issues and it won’t let you make the right decision, you can use Mental Models to get out of the misery.
Let’s get you Started with Mental Models
Not every problem can be solved with one solution, as every problem has different complexities and requires different approaches. Most data scientists tend to solve a problem with one old traditional approach and keep their eyes on deep issues while the solution is at the front.
With Mental Models, we can do problem-solving with several methods that are effective, fast, and give you output.
Mental models revolve around four basic categories:
- Thinking (Brainstorming)
- Problem Solving
Below we’ve given examples of each category; some examples may resemble your scenario.
Note: These categories are not perfect, and you may find a need to input your understanding according to your situation.
A New Problem
Suppose you’re a Data Scientist working for a company, and you have to make decisions and lead a team of other junior data scientists.
Now, you got emails about the new problem:
1st Problem: You have to deal with a new company that offers mental health treatments to people through music. They created an app where a person has to select his/her mood, and the app will play something according to the music with the help of AI.
The issue is that the user can’t reach the desired mood goal. (the startup company also did scientific research, but still, models are not working)
2nd Problem: A bank wants to predict the amount people will take from the ATMs in 30 days. They want to get the amount daily.
Some Effective Thinking Mental Models in Data Science
Some of the best thinking mental models are below:
This mental model is the most effective and useful because, in this mental model, we display the system visually. When things are in front of us visually, we can mark the linkages between each part. So, the first thing we recommend when dealing with a new problem is to design a concept map.
“Concept mapping has been shown to help learners learn, researchers create new knowledge, administrators to better structure and manage organizations, writers to write, and evaluators assess learning.” – Joseph Novak & Alberto (Creators of the Concept Map)
To create a concept map, you need to follow four major steps.
- Formulate a Problem Question
The first step is to know the big questions we need answers to. These questions usually represent the point where a problem is occurring. You may ask how your app works. What are the core components of your algorithm? What is the ecosystem of the app?
- Identification of Key Entities
The next step is to identify the key entities. For this purpose, you must create a list of key entities that may be linked to a problem, such as algorithms, people, processes, and so on…
There is no hard-and-fast rule about the number of key Entities, as you may get more or fewer entities depending on your problem.
- Sorting the Entities
Now that you have done with the identification of key entities, the next step is to sort them according to their importance. It will help to create the concept map by hierarchy.
- Map Outlining and Filling
In this step, we add entities according to the hierarchy and then link them together with arrows to represent the direction of actual actions.
We got a flowchart-type concept map, and now you can easily understand what the startup is doing and what we can do about the problem.
With the help of the Iceberg Model, we can notice the problems and reasons that are not visible. It helps to identify the complex side of the problem in Data Science.
The iceberg model further has four components:
Events: What’s happening? What is the problem? What is the end goal?
Patterns: These represent previous trends and historical data.
Structures: These represent the interconnection between different parts. I.e., How do parts get connected?
Mental Models: Is there any Mental model? What kind of mental modes are present?
If we talk about the bank’s problem, it’s very clear that you have to gather previous data and come up with a prediction. But, with the Iceberg model, you get to know the hidden parts of the request.
→ Why do they want to predict the upcoming 30 days?
→ Why do they want the prediction daily?
→ What use Case?
So, the next step is to ask these questions from the bank. On asking the bank, we got the answers…
→ It will help them to fill the atm before it goes empty.
→ 30 Days are the most convenient
→ They can schedule weekly atm filling.
Now, we can check patterns, such as:
Salary Days – When people get paid, they use ATMs
Location: ATMs located in different places are used less or more. In urban areas, people use more ATMs
Weather – When the weather is not favorable, people do not go to the ATMs
Event – During the wedding season or Christmas, or black Friday, people use ATMs more
Reinforcement Feedback Loop
Suppose you shout on the microphone, which is placed next to a giant speaker; you will realize that you should not have done that, right? A Reinforcement Feedback loop is a loop that amplifies its effects.
Let’s take an example of that music app. If a person is depressed, and he or she selects depressing music (because they want to listen to this type of music), they will get more depressed. Sometimes, the music you choose can harm you, and selecting music that may make you feel better is sometimes daunting.
Let’s take another example if you’re deciding with a team; If there are more people on your team, it will be more difficult to make the right decision. (You may make a decision that may lead to chaos). Of course, people will follow the person they like, and at the time of the final decision, the signal amplifies the loop.
So, it’s crucial to fight these scenarios. For this purpose, we can ask our team members to write their stance on a specific problem anonymously. Further, discuss the opinion received from the team and make a better decision.
Balancing Feedback Loop
As the name suggests, it is a loop that balances or counters the RFL. You can witness BFL in many problems or systems because it’s usually present with RFL. While making decisions in your data team, there will be one person that will handle the balancing.
Now let’s take an example of the music app; we will ask them to check the current mood of the user and what mood they seek after using the app.
For example, if a user loves self-destruction, he/she will listen to music that will amplify his state. Here, we need BFL to give a user music that will benefit his/her mental health.
We will need more data and study to make it more helpful for the user.
Problem Strikes Back
The startup listened to our suggestions and implemented everything. Now, they want to improve their algorithm for users to increase the quality. To improve it, one wants to add surveys for more data collection, and the other wants to include biometric data tracking.
Some Effective Techniques for Decision-Making in Data Science
We’ve highlighted some of the techniques for Decision-Making in Data Science.
Six Thinking Hats
In this technique, we try to solve problems and think about the problem from different angles, and drive the right path to make a final decision. You can do it alone, but we recommend a team.
Let’s take you to this interesting technique. In this method, we assign different hats to each individual from the team:
Green Hat: Think of ideas in many directions with creativity.
Yellow – List out the benefits of a specific decision
Black – List out the cons of a specific decision, and check the vulnerability of a decision
Red – Check Emotional Feelings.
Gray: Pay attention to the data in a rational way.
Blue: This one is the most important. Blue hat people maintain the decorum and moderate other hats. ( We recommend a team leader to be the blue hat)
The Cynefin Framework
With the help of Cynefin Framework, we can understand the problem and design a set of steps to solve it. There are five categories of this Framework.
Clear – In this Category, things are mostly clear, and the problem is easy to understand.
Complicated – In this category, things are a bit complicated, and a person needs to ponder the solutions to problems
Complex – In this category, things are complex and require proper investigation and analysis to fight the problem.
Chaotic – In this category, systems are not stable and require stability to deal with them.
Disorder – In this category, we don’t know the category of a problem. We usually split the problem and then categorize them.
Did you notice that our bank problem was a complex category? But, still, we have to solve the problem, whether it’s straightforward or complex.
Data Scientist Returns
We’ve already found a solution to the problems, but this is not the end. We still need to think of ways to make it smooth and find more options.
We will make some decisions and then forward them to the data team.
Some Effective Problem Solving Techniques in Data Science
We’ve enlisted some of the techniques of problem-solving in Data science:
This technique breaks down a problem into micro components that can not be divided further. And implement the basic truths to come up with the solution.
To get the basic truth, we ask a series of questions until we get the basic truth. This method is called the Socratic Method.
Let’s take an example of the bank where we asked multiple questions and got the basic truths:
→ We need to refill the ATMs
→ ATMs at specific locations require more frequent filling
→ During events or holidays, we need to refill ATMs more often
When we get the basic truth, we model the problems and frame regressions, classification, aggregation, optimization, and survival.
The Inversion Approach is used as a pessimism. This approach considers the worst-case scenario by analyzing the problem from multiple angles.
In this approach, we consider that implementation has failed, and then we find the reasons for failing, our mistakes, and things we should have considered before. It will help to make the decision foolproof.
Sometimes, when everything is in control and everyone is doing fine in finding the solution to a problem, you may notice one person from the data team is not working properly, and you may notice your intern is off track a little bit.
It’s natural that an intern may take pressure from the leader, mess up everything, and act aggressively. That’s where you need communication skills to hear the issues of your teammates.
If you’re good at communication, you will kill your problems like a pro!
Some Effective Techniques for Improving your Communication Skills
There are a few techniques you can use to improve your communication skills. Such as,
- Assertive Communication
- Listen Actively
- The Minto Pyramid
SBI stands for Situation Behavior Impact. With the help of SBI, we can get better feedback and eliminate emotions. When we eliminate emotions, we get a clear view of a problem.
As it is said, don’t let emotions come your way. So, we park emotions aside and give negative feedback to improve the situation.
We implement the SBI technique in 4 steps…
Situation: The feedback starts with a situation.
Behavior: You notice the behavior of the situation so that you may address it.
Impact – The third step is to address the impact of the situation
Intent – See the person’s intention and ask about further things (Why did he do it? What was his intention?)
When you’re aware of all the things above, it’s easy for you to make him/her understand.
The Minto Pyramid
In Minto Pyramid, I re-arrange a message so that we get a conclusion and then arguments that will support the conclusion. It will enable us to make a message clear and easy to understand.
Now let us take you to the implementation of the Minto Pyramid. We will follow the Minto Pyramid format of writing a message. So, we will write the conclusion part short and the other part in detail.
Nobody has time to read your long messages written without any formatting. Clear and concise messages will hit the spot.
A good data scientist is not the one who can write good code and develop like a pro. There are hundreds of problems a data scientist faces while working on a specific project, and to fight that problem, you must know how to do critical thinking.
This article explains the tips and tricks of implementing mental models to improve your critical thinking and find an optimum solution to a problem.
By following mental models techniques, you can win any project like a pro…
want to learn how to algo trade so you can remove all emotions from trading and automate it 100%? click below to join the free discord and then join the bootcamp to get started today.