After reading this post you will know:
- What Decision Intelligence is
- When to apply it
- How it interacts with Data Science
Okay, let’s be honest: sometimes you cannot see the (scientific) field, due to all these vague terms, that are thrown around, like there is no tomorrow. Let’s have a look at Data Science:
If you search for Data Science, you will end up with alternative suggestions such as Business Intelligence, Machine Learning, Artificial Intelligence, Big Data, Data Analytics, etc. And I’m only getting started: Deep Learning. There, I wrote it!
Similarly for Decision Intelligence. Terms such as Decision Analysis, Decision Theory, Behavioral Economics, Business Intelligence (seems like this one needs a bit more attention), Decision Quality, Decision Engineering, Behavioral Decision Theory, etc. – you probably get the pattern – can lead to confusion.
Fortunately, in Decision Science, similar to Data Science, the different terms are in some way related. This is of course true for every scientific field, if you simply look hard enough.
So let us start with the most obvious question:
What is Decision Intelligence?
To put it quite bluntly: I found around 14 “definitions”. Some are good, others are there, and one is infuriating. Though I must admit, that the most infuriating post was the one with the best insights in regards to reader comments. If this was the plan, I am quite so very much intrigued and would like to hear more about the decision process.
Back to the point!
There are three definitions that I think, give a very good impression of what Decision Intelligence is and what we expect it to be capable of.
“Fit between Intelligence Requirements of an organization/individual and the Intelligence Processing Capacities of the same.”
Roger Moser – Decision Intelligence PVT.LTD:
https://www.decision-intelligence.org
“A framework that brings multiple traditional and advanced techniques together to design, model, align, execute, monitor, and tune decision models and processes.”
Senior Analysts at Gartner:
https://medium.com/decision-automation/what-makes-decision-intelligence-a-better-framework-for-decision-making-models-8348a72c3959
“Bring together the best of applied Data Science, Social Science, and Managerial Science into a unified field that helps people use data to improve their lives, their businesses, and the world around them.”
Cassie Kozyrkov – Head of Decision Science at Google: https://towardsdatascience.com/introduction-to-decision-intelligence-5d147ddab767
There is a lot hidden in these three quotes. Let us have a look at the first one:
“Intelligence Requirements” and “Intelligence Processing Capacity”, to put it very simple, are descriptions for the necessary resources (time, money, information, staff, etc.) you need, to make good decisions. The former is usually used to define what is necessary and the latter is the collective phrase for the resources you actually have. Though it does make a difference, the most important part to remember would be the phrase: resources for decision making.
The second quote has one interesting statement: “…multiple traditional and advanced techniques…”. On YouTube you can find a video done by the Strategic Decision Group which focuses on Decision Science for Data Science (which we will further discuss later) and during the presentation there is a slide which states various Theories, Best Practices, and Tools that are needed/should at least be known, to be an adequate decision professional. You can see the table below.
I do get that Table 1 is a lot and do not worry, I will only shortly write about the Theoretical Foundation in this post. The points stated in the lowest row of Table 1 include interesting and important fields such as Probabilistic, Finance, Psychology, and Business (I may have exaggerated with that one). This leads to the conclusion that Decision Intelligence is a very diverse field which demands a lot. And also gives you a lot of possibilities. Like one of those unwritten rules of life: “You get what you pay for.” Except when it is about interest rates. Then you just pay without getting anything back…
Definition Nr. 3! This one uses our insights of the two definitions before and adds Data Science to the equation. Before you roll your eyes and think something like: “Nowadays everything has something to do with Data Science!” hear me out: “Deep Learning!” 😉
But in all earnest, the need for Decision Science in the area of Data Science is given. AI expectations are rarely met, which can leave a bad taste in the mouth of businesses. Additionally, the strength of Machine and Deep Learning lies in the fast execution of repeatable tasks. In other words: tasks that must be repeated all the time can be learned by a model and then easily executed, giving a human the chance to focus on work that demands critical thinking, social interaction, and is non-repetitive.
Before we move on to our next segment of the post a quick summary: Decision Intelligence is defined as a field that focuses on decision making by utilizing different (scientific) fields (you all know why scientific is in brackets).
Why is Decision Intelligence necessary?
You may like the concept of the Gartner Hype Cycle or not; Nonetheless, it is a good indicator of what topics are hot and which are not. Sorry Cognitive Computing!
What makes the 2020 Hype Cycle for AI interesting to us, is the mention of Decision Intelligence. DI is currently at the rise and situated next to “Smart Robots” and “AI Developer and Teaching Kits”. This is a big win because it increases the visibility of DI.
Of course the argumentation doesn’t end here. I will again use a quote, because everything sounds smarten when it is a quote:
“Discrepancies between sophistication of organizational decision-making practices and the complexity of the situations, in which those decisions need to be taken.”
https://research.aimultiple.com/decision-intelligence/
Though the quote only focuses on organizations, it can also be applied to people. Either way, what the quote states is that we tend to compare apples and pears, when it comes to the capability of making sophisticated decisions, and the complexity of the systems (environment, economy, educational institutes, sugar consumption, etc.), in which these decisions are made.
Lorien Pratt uses a great analogy (which I may have embellished a bit) for our current state of decision making:
We used to be like fish in a pond. Everything was foreseeable as possible. You knew your peer-group, you knew about the area, and that was it. Nothing beyond the borders of your “home” was any of your concerns. And then suddenly the pond opened up and became part of the ocean. New animals arrived and left. Your entire world changed and with that all the decision making became unpredictable. Whom could you trust and where could you go for food and safety?
My intention with these two statements is not to frighten you, quite the contrary. I absolutely do understand if people feel overwhelmed and simply go with the easiest option, because there are way too many. Side note: have a look at the comedy series The Good Place. They focus on concepts such as decision making. Totally worth a try.
And now we finally get to the main point of this part: DI can be used as a solution for this problem.
Image 2 can be interpreted in the following way: The rows represent the amount of data. Top row means little to none, whereas the last row is synonymous with Big Data. The columns on the other hand represent the type of data. Examples for hard data could be logging files, regarding user preferences in an online shop. Soft Facts, on the other hand, can be described as experience, intuition, rules of thumbs, etc. With that in mind, let us have a look at AI and those three arrows. Let’s start from the bottom.
Due to the massive amount of data available, AI can be used rather straight forward. For Hard Facts well known algorithms such as Neural Networks and the k-Nearest Neighbor can be used. Soft Facts can be put to use by e.g. using Knowledge Graphs or Decision Trees. Examples for this section would be driverless cars, automatic tagging, etc.
The middle section – Operational Decisions – benefits by insights generated with the help of AI. This is the domain of Data Analytics. I do not want to go to deep into this rabbit hole, because this is not the point of this post. Simply know that there are four types of Data Analytics:
- Descriptive: What is happening in my business?
- Diagnostic: Why is it happening?
- Predictive: What is likely to happen?
- Prescriptive: What do I need to do?
As a rule of thumb: these are decisions an organization needs to make once in a while, but not on a daily basis. Examples would be: Marketing Plans, Supply Chain Risk Management, and unscheduled Audits.
The final category: Strategic Decisions. This is where humans are at their best, in comparison to AI. Strategic Decisions are decisions that take a long time to plan and are made so seldom that little to no data is available. E.g.: Organization Re-branding, Acquisition of a competitor, Organizational Capability Transformation, etc. In this case, AI can, if at all, only be used to gain some insights; Hence, fields such as Managerial Studies, Psychology, and Descriptive Analytics can be used to get inspired. Applied one could say: “Formulate questions, which need/should be answered, before a decision is made.”
Now we finally have an understanding of what DI is capable of and to what extend: We can use it to automate decisions that are easy to learn, if enough data is available. We can also use it as a framework to gain new insights and help us choose, maybe not the best, but at least a more optimal choice than random.
With this being written, let us move on to the third part of the post.
Okay, I get it! Can I finally go back to my AI models?
Bare with me, because we are not done yet! There is one more issue: the execution of AI projects. Yes, I know: I already wrote about when and how AI should be used the best, and I also stated, that AI projects do not deliver on their expectations, and that there is more to AI than Deep Learning. What else would I like you to read?
Cassie Kozyrkov has one or two things to say about that! If you do not have the time or simply just don’t feel like it (though video Nr. two is really recommendable!) here is the gist: Make decisions, to make AI, to make decisions. Don’t start using data like Xanax for the sake of using it. Use it for a purpose!
To put this into a graphical representation:
The Venn-Diagramm can be interpreted as the following: Find the common ground of what is possible with data and what is useful for the user. Because at the end of the day: the user pays! In other words: it is important to have people in an organization that speak data, know the importance of User Experience, and also know how to talk to the C-Suite. Easy-peasy! Right! Right?
I will end this section here, because there is, besides the videos by Cassie, a good post by the lead Decision Scientist at Instagram. The thing I want you to remember though is that Decision Scientists and Data Scientists are closely related, make amazing teams, and can help you get a lead in the game.
With this being written, let us have a look at the final part of this post: “Whats up next?”
Okay, you made your point! So what’s next?
This one is easy and straight forward: have a look at different Decision Intelligence sources. Below you can find some potential sources that I consulted for this post.
Blog Posts:
Books:
- Foundations of Decision Analysis by Howard & Abbas
- The Black Swan by Taleb
- How to Measure Anything by Hubbard
Videos:
This leads to the end of this post. If you liked this post, didn’t like it, and/or have a book recommendation, please leave a comment, subscribe to the blog, and share this post. I am always happy to learn new things and improve myself.
DISCLAIMER: The visualizations presented in this graph do NOT represent quantitative data. They are merely used for argumentation purposes and therefore are not scientifically sound! For more information & a good laugh, have a look at the book Calling Bullshit by Bergstrom & West, Chapter 7: Data Visualization.
Nice introduction into Decision Intelligence! …it’s a job, right?
Thanks! Yes, it is. Though it needs to be stated, that job titles vary. San Francisco State University offers an interesting document:
https://cob.sfsu.edu/sites/default/files/career_services/Decision%20Sciences%20Final_0.pdf