Do you ever believe AI is being leveraged to separate people from their hard-earned income? Consumer platforms, like Amazon and others, have an economic incentive to use AI to maximize platform profitability. [i] We will show you how to implement your own personal decision algorithms. This is the answer for overcoming the consumer platforms and their money-hungry AI.
Data is everywhere. In just the last 30 years or so, our world has relentlessly flipped from the data-scarce world of old to the data abundance world of today. [ii] We are literally drowning in data.
Unfortunately, our own brains are not yet prepared for the new data-abundance world. It is not our fault! It takes thousands of years for natural selection to do its evolutionary job. Our brains are still geared to making decisions in pairs, like a "fight or flight" decision. In the meantime, we need to adapt. This adaptation means developing personal decision habits more suitable for today's more complex decisions. Our own neurobiology is not yet naturally equipped to handle today's common, complex decisions.
In this article, we show you how to do it. We will show you how to be a personal data scientist without knowing statistics! We are going to show you tools that do all the mathematical heavy lifting in the background. We use a common pet purchase as an example, but the approach could be applied to almost any purchase. We will show you tools like Definitive Choice that pair with how we naturally make data-scarce decisions and then convert them to the data-abundant world.
About the author: Jeff Hulett is a career banker, data scientist, behavioral economist, and choice architect. Jeff has held banking and consulting leadership roles at Wells Fargo, Citibank, KPMG, and IBM. Today, Jeff is an executive with the Definitive Companies. He teaches personal finance at James Madison University and provides personal finance seminars. Check out his new book -- Making Choices, Making Money: Your Guide to Making Confident Financial Decisions -- at jeffhulett.com.
Cover photo credit: The Curiosity Vine and Freepik.com
Table of Contents:
Introduction
Create statistical habits without being a statistician
A pet perspective
The connection to personal finance
Resources for your personal algorithms
Personal algorithms best practice
Conclusion and notes
2. Create statistical habits without being a statistician
Robyn Dawes and Daniel Kahneman are well-known behavioral psychologists and behavioral economists. Dr. Dawes was one of the first to observe that people can get much of the power of statistics using an approach he called "improper linear models.” [iii] For our discussion, we call Dr. Dawes’ ground-breaking research “personal algorithms.” He and others showed how personal algorithms can be just as good as statistical algorithms, and certainly better than algorithms suffering from overfit or a lack of accuracy resulting from uncertainty. Dawes and other researchers showed how a set of independent variables, likely correlated to the dependent variable, with judgmentally set coefficients will likely perform quite well. The Nobel-winning behavior economist Daniel Kahneman [iv] said:
"The immediate implication of Dawes’s work deserves to be widely known: you can make valid statistical predictions without prior data about the outcome that you are trying to predict. All you need is a collection of predictors that you can trust to be correlated with the outcome."
Thomas Saaty is the inventor of the Analytical Hierarchy Process (or "AHP") and a decision scientist. AHP, among its other benefits, is a decision method that transforms how people naturally make good judgments in pairs. Owing to our evolutionary biology, people are genetically predisposed to make highly accurate comparative judgments, called "pairwise comparisons." AHP leverages our natural pairwise ability to make more complex multi-criteria, multi-alternative decisions common in everyday life. [v] To connect the dots, Dr. Saaty's AHP process enables Dr. Dawes' personal algorithms. The essential step for implementing a personal algorithm is to weigh the predictive criteria correlated with the decision outcome. [vi] AHP renders highly accurate predictive decisions utilizing a judgmental process available to any person or group. By the way, both Dawes' and Saaty's groundbreaking work was published in the 1970s. Their work has increased in relevance as needed for today's more data-abundant world. In the resources section later in this article, we suggest Definitive Choice. Definitive Choice implements AHP to help develop your personal algorithm and make common multi-criteria, multi-alternative decisions.
The research shows that people can make personal algorithms that are just about as good as statistical models developed from rich data sets. These personal algorithms have a few other tremendous advantages:
Personal algorithms are unique to you. It allows for your incredible originality and that which makes you special. It does not force you to be the average of everyone else.
Personal algorithms are adaptable. As you change, the personal algorithms change to your changing beliefs. Personal algorithms improve as you improve.
Personal algorithms separate new information from the undue influence of situational framing. Much work in psychology shows how people are impacted by situational framing. [vii] Behavioral economists' central thesis demonstrates that people are inconsistent in making decisions depending on how the situation is presented or "framed." For example, you may be more likely to buy a pet after a positive interaction with a puppy than a negative interaction. Sometimes, new information provided as framed in a new situation is helpful, sometimes not. A properly implemented personal algorithm will help accurately separate new information from the undue influence of situational framing.
Aligning incentives. In today’s hyper-competitive consumer platform world – think of big consumer platforms like Amazon, Netflix, AirBnB, and others - their incentives are not necessarily aligned with helping you make the best purchase decision. Not surprisingly, big, stock-held consumer platforms need to drive quarterly revenue to satisfy their investors. Thus, their AI is geared toward getting you to buy as much as possible. Personal algorithms help you clarify “what is important to YOU” instead of being persuaded into buying something more important to the quarterly investor report.
Personal algorithms are the answer to overcoming the consumer platforms and their money-hungry AI. This approach shows you how to harness your own HI 😊 or “Human Intelligence” to thrive in today’s data-abundant world. Next, we demonstrate personal algorithms by using a standard pet decision example. This can be applied to almost any purchase decision!
3. A pet perspective
In 2023, according to a Forbes Magazine article, there are about 87 million pet owners in the U.S. According to the ASPCA, the average pet owner spends nearly $1,400 annually on their furry pal. Thus, the total estimated annual cost of pet care is a whopping $122 Billion.
To start... essential to evaluating your pet purchase decision is to understand your core motivation. In this case, the question is - "What is important to you about owning a pet?"
Evaluating pet owner benefits:
The essential pet criterion weighs the degree to which your and your family's emotional support is provided for by the pet.
According to Psychology Today, the essential reason people buy pets is because of the emotional support provided by their pets. The summary reasons are:
Pets attend potently to our brain's social circuitry.
Pets are pure. The pets' innocence is inspiring.
Pets only know and breathe connection.
In short, pets provide their owners emotional support in exchange for the pet being cared for by the owner. This is also known as love.
Besides the love benefit, according to the VCA Animal Hospitals additional pet ownership criteria are:
1. Home environment (size):
The size of your home and yard, as well as your lifestyle, will play a significant role in determining which type of pet is right for you. For example, if you live in a small apartment, a large dog may not be the best choice.
2. Temperament:
The size and temperament of the animal should be compatible with your family’s lifestyle and needs. For example, if you have young children, you may want to choose a breed that is known for being gentle and patient.
3. Training and exercise needs:
Different pets have different training and exercise requirements. It’s essential to choose an animal that you can commit to training and exercising regularly.
4. Physical characteristics:
The physical characteristics of the animal should also be taken into account when selecting a pet. For example, if you have allergies, you may want to choose a breed that is hypoallergenic.
5. Lifespan:
It’s important to consider the lifespan of the animal when selecting a pet. Some pets, such as birds and reptiles, can live for decades.
6. Compatibility with other pets:
Other than pet benefits, prospective pet owners need to consider pet costs.
Initial and recurring costs:
Before choosing a pet, it’s essential to consider the financial commitment involved in pet ownership. This includes the cost of purchasing or adopting a pet, as well as ongoing expenses such as food, veterinary care, and grooming.
Properly evaluating and weighing criteria is weirdly challenging. Decisions such as a pet purchase are deceivingly complex. This is owing to the proper weighing of the criteria set and then applying the criteria to all the pet alternatives to make a cost-benefit decision. Also, Adam Smith teaches us that criteria impacting us further in the future are even more challenging to properly integrate into the decision. [viii]
Comparing Artificial Intelligence to Human Intelligence:
Next, we illustrate how AI, the AI platform - such as a big consumer platform, and your HI or personal algorithm interact.
First, for your pet criteria, we simplify the algorithm to allow the love and social connection component to be the driving criterion. You could certainly add other criteria, this assumption just makes it easier to illustrate. In general, those dots receiving higher benefits - as in groups 1 and 4 - are more socially oriented in personality, like an extrovert. Those receiving lower benefits - as in groups 2 and 3 - are less socially oriented in personality, like an introvert. We also demonstrate how long-term pet ownership is relatively expensive. In the next section, we show how the data supports a long-term pet ownership opportunity cost of about $1 million to your retirement. Finally, we assume that there are much lower marginal cost substitutes for social-emotional support, such as church groups, clubs, family connections, private counselors, and many others.
This graphic shows the view via the AI and the HI algorithm.
You & what the AI sees: The AI sees you as one of many dot data points. Based on the data, the AI understands there is a non-linear, parabolic-like relationship between pet owner benefits and the cost of those pets. The AI determined this by using neural networks, decision trees, or other statistical modeling techniques. Upon determining and then executing the best statistical model, the AI will render the mathematical model of the parabola so it can predict which dot data point you are in the cost/benefit space. The math specifies the orange, parabolic-shaped line that fits the darker gray dots. Also, more advanced AI's regularly update the statistical model as new data becomes available. This updating process is called "Machine Learning."
Next, we show how the AI platform company interprets the results of the AI.
You & what the AI platform company sees: The platform company sees an opportunity to convince you to move to the right on the cost axis. This is because the consumer platform makes more money when you spend more on a pet. So, the AI platform company uses incentives and marketing to convince you to purchase a pet or pet supplies. The AI platform company needs to make its numbers next quarter! Based on the graphic, the AI platform company will attempt to convince you that you should locate in the high-cost pet space, such as dot number "3" or "4". To be fair, the AI platform has a complex set of motivations tugging at their self-interest. These motivations include trade-offs between short-term profits, long-term profits, customer needs, employee needs, broader community needs, and others. However, generally, a stock-held company is very sensitive to quarterly profits. The primary way an AI platform company makes next quarter's profits is by selling to you TODAY.
What you see: You are one of these many data points. You may do research or talk to your friends, but in general, you are not aware of the full cost/benefit space. Broader data knowledge is the AI's advantage. If you have a clear understanding of the benefits of a pet, what economists call "utility," then you can make a great decision about how or whether to buy a pet. Your unique HI advantage is:
Your clear, unique benefit knowledge and
Your power of choice. The money is still in your bank account. You have the power of "no."
The graphic highlights 4 narratives, numbers from "1" to "4", as examples. Your pet story will be unique to you. For example, if you know you are the "1" data point, you will confidently pass on a pet. Also, if you have a clear understanding of your pet benefits, you will actively seek to avoid becoming data point "3".
Interpreting the graph - who wins?
You WIN – in quadrants 1 and 2, if you do NOT buy a pet; and in quadrant 4 if you do buy.
AI WINs – in quadrants 3 and 4 if you buy.
The ONLY overlapping quadrant where both the company AND you win is 4.
Please note: Be careful assigning a judgment such as "The AI is "good" or the AI is "bad." The point is, AI is just a tool that can be used for good or bad. It would be like calling a hammer or a screwdriver "good" or "bad." The assignment of "good" or "bad" is related to your (the human's) motivation or preference set alignment with the AI platform company. The higher the misalignment of motivation, the higher the likelihood the platform company's A.I. will be bad for you.
However, if you DO NOT have a clear understanding of your benefits, you will be susceptible to AI-informed influence by the platform. The consumer platform will do its best to move you to the right on the horizontal cost axis. Not having a clear understanding of your utility exposes you to the higher costs and lower utility associated with economic discrimination. [ix]
Next, explored is how to understand long-term pet costs. Then, a personal algorithm app is suggested to help you develop your HI and make the best decisions. We will continue to use the pet example, but this approach may be applied to any personal finance decision.
4. The connection to personal finance
In the context of personal finance, our "payment hierarchy" is an essential consideration when thinking about pets or other long-term financial commitments. A pet is like a fixed expense - such as a car loan, a mortgage, or other contractual debts. In my book, we discuss the need to "pay yourself first" and to avoid a well-intended expense devolving into "sludge." [x] In the context of the pet, the challenge is to properly categorize the pet decison into one of three categories:
an investment,
a need-based expense,
or a want.
Getting the payment hierarchy correct and operationalized is essential to long-term personal finance success. Also, implementation enablers called "commitment devices" help ensure your payment hierarchy compliance.
Depending on where the pet falls, will drive the decision process and long-term outcomes. Next, we discuss the challenge of categorizing the pet's utility and then the financial implications of the pet acquisition. Also, because of the pet's nature, the pet will generally be prioritized at the top of the payment hierarchy AFTER THE PURCHASE. As such, getting this decision correct PRIOR TO THE PURCHASE is important as the pet decision is challenging to unwind.
The challenge is forecasting your utility over the likely expense time period. Let's assume a pet is a twelve-year commitment. Are the emotional benefits you and your family will receive worth the long-term costs? Let's face it, a fluffy puppy is as cute as a button. Next are some utility-clarifying questions:
Can you imagine that puppy as a full-grown dog?
What about your family? As they grow and change, are they likely to achieve the same benefits they receive today?
What is the benefit of the pet's emotional support to you and your family? Are there alternatives for that emotional support?
Are you committed to paying the long-term costs (money and time) to care for the pet?
Finally, how are those benefits prioritized relative to the funding substitutes? - like providing more resources for your retirement or other long-term priorities?
While forecasting our utility is challenging, forecasting the cost impact is achievable with a simple model. We know average pet costs are $1,400 per year. We also know the average age of a dog is 12 years. Let's say you acquire a pet in your early-to-mid 20s. In this example, we will assume you will keep the pet for its full average life. Also, we will assume some people will like pet ownership so much, that they will buy a second pet for an additional 12 years in the mid-to-late 30s. So we have 2 scenarios: a "1 Dog" for 12 years scenario and a "2 Dog" for 24 years scenario.
Source: Pet Ownership Model
This graph shows the opportunity cost of diverting the pet ownership costs from retirement savings. The point is to put a hard value number on the pet ownership trade. Whether you use these resources to solve world hunger, your own retirement, or pet ownership is left up to you!
The assumptions for our simple model are that the $1,400 / year cash flows are invested in investment funds receiving an annual return of 10% over the pet owner's life, based on the two 12 or 24-year pet ownership scenarios. Retirement is assumed at 65 years old. Thus, in the 2 dog scenarios, the opportunity cost of pet ownership is:
So, in round numbers, the lifetime cost of a pet is about $1 million. To be clear, this does NOT include the indirect costs of your time to care for the pet. The modeled costs are food, medical care, and other direct costs. This should help you make the tradeoff decisions for your own utility and benefits.
So, before we started, a common belief is “I want to buy a pet!” Now, our eyes have been opened. Pets can be very expensive and there are less expensive alternatives to consider. Now, your more informed perspective is “I want to buy a pet in a SMART way.”
5. Resources for your personal algorithm
Now, it is time to determine your baseline algorithm for deciding the best pet for you. Notice, that this is geared to you, not someone else. Also, you can use Chat GPT to help you decide your criteria. For example, for this criteria, I simply went to Chat GPT and asked “What are the top criteria for buying a pet?” The next step is to create a weighted personal algorithm that is perfect for you.
Definitive Choice is an app decision solution to help you understand your own self-interests and actions on almost all life decisions. The app can help you and your family buy a pet or consider other alternatives.
It provides a straightforward user experience. The number-crunching occurs in the background by time-tested decision science algorithms. It uses a proprietary "Decision 6(tm)" approach that organizes the preference criteria (what is important to you?) and alternatives (what are the choices?) in a series of bite-size ranking decisions. Since it is on your smartphone, you can use it while you are curating data to support the decision. It is like having a decision expert in your pocket. The results dashboard provides a rank-ordered list of recommended "best choices," tailored to your preferences.
Also, Definitive Choice comes pre-loaded with many decision templates. You will want to customize your own preferences (aka criteria) and alternatives, but the preloaded templates provide a nice starting point.
Using decision process solutions enables DECISION A-C-T:
Accelerated: faster, less costly decisions. It enables a nimble decision environment.
Confidence-inspired: process causes people to be more confident in the decision, increasing buy-in, and decision up-take.
Transparency-enabled: reporting, documentation, and charts to help communicate the decision.
6. Personal algorithm best practices
Next are a few ideas to help you make the best decisions! This will help you build the best personal algorithm with easy-to-use tools and apply your personal algorithm to potential alternatives. We use the pet example, but the essential element is to get in the habit of making consistent, repeatable decisions in today’s data-drowning, data-abundance world.
Clear Criteria Definition: Clearly define each criterion for your "what is important to you” utility model. These are the benefits you receive from the acquisition of the objective. In this example, the objective is a pet. A ChatGPT question like “What are the top criteria for purchasing a __________?” is a great way to confirm your criteria definitions. Each criterion should have a series of dot points to clarify its definition.
Build your own criteria: Definitive Choice provides typical criteria as a starting point. We encourage you to develop your own and update the list. The same ChatGPT question is helpful for confirming your criteria.
Reduce criteria to most essential: Generally, you do not want more than 5 criteria. So, subtract the criteria you know are of no or little value to you. For example, if you do not already have a pet and do not anticipate having multiple pets, then the "Compatibility with other pets" criterion may be deleted.
Ask for help: Definitive Choice allows you to invite others to participate in the decision by sharing their own utility model. This is especially helpful for those who are first-time buyers. The invited person should be a “Conscience Person” and someone who has experience with the purchase. A conscience person is generally someone you do not want to disappoint… like a parent or a respected friend.
Consider costs separate from benefits: You can include cost as benefit criteria, but you run the risk of double counting costs. We recommend separating cost from benefits and considering the tradeoffs between the costs and benefits as separate dimensions.
Criteria should be independent: A natural tendency is for people to mentally combine separate criteria categories. It is essential to BOTH define each criterion definition AND to ensure each criterion is independent of the others. This can be a little tricky. For example, for a pet, you want to consider a pet with the appropriate size to fit the environment and the size and temperament to fit your family's needs. In this case, size can be independent as long as it is considered separately from the environment and the family members' needs. However, if those 2 criteria are always the same, then you should combine the criteria into a single criterion. For example, if you have an apartment requiring a small animal and you have young children that require a small animal, then combine these 2 criteria into a single “Size” criterion. Perfectly correlated criteria run the risk of double counting or overweighting. [xi]
7. Conclusion
Whether buying a pet or other significant financial decisions, using a personal algorithm is essential for making the best decision. You can reap the benefits of statistical analysis without being a statistician! The summary recommendation is:
Create a consistent, repeatable decision process you can use across all your decisions.
Use a personal algorithm to implement your consistent, repeatable decision process.
Definitive Choice is a very good starting point to implement your personal algorithm.
The idea is not to either encourage or discourage pet ownership. The idea is to provide a decision process so you have conviction in your confidence about the pet ownership decision.
Notes
[i] Hulett, Top 6 reasons why Personal Finance success starts with choice architecture, The Curiosity Vine, 2023
[ii] Hulett, Solving the Decision-making Crisis: Making the most of our free will, The Curiosity Vine, 2023
[iii] Dawes, The robust beauty of improper linear models in decision making, American Psychologist, 34(7), 571–582, 1979
[iv] Kahneman, Sibony, Sunstein, Noise: A Flaw in Human Judgment, 2021
[v] Saaty, How to make a decision: The analytic hierarchy process, European Journal of Operational Research, Volume 48, Issue 1, Pages 9-26, 1990
[vi] Korhonen, Silvennoinen, Wallenius, Öörni, Can a linear value function explain choices? An experimental study, European Journal of Operational Research, Volume 219, Issue 2,
2012, Pages 360-367
[vii] Tversky and Kahneman, The Framing of Decisions and the Psychology of Choice, Science, 211, 453–458, 1981
[viii] Adam Smith published The Theory of Moral Sentiments ("TOMS") in 1759. This was the prequel to his even more famous book The Wealth of Nations. (1776) In TOMS, Smith describes the 4 sources of moral approval. These sources are appropriate for market interactions or more general life interactions. Think of these sources as the approval between agents for some interaction. The 4 sources of moral approval is part of a naturally occurring recursive process between decision agents. This process operationalizes what Smith calls "The Invisible Hand." This interaction could be buying a good - such as buying a pet.
The four sources of moral approval (TOMS 326-327.16), are:
Approval of the giving agent. (supply or pet seller)
Approval of the receiving agent. (demand or pet buyer)
Approval of the interaction environment (These are the rules or norms governing today’s interaction environment, such as the criteria impacting the pet or pet owner's environment. It could be the local animal control laws or building rules for an apartment. It could also be various local customs about pet ownership.)
Approval of the unseen (long-term impact of 1, 2 & 3. This is the long-term impact of pet ownership, including the longer-term questions asked in the article and the long-term financial impact.)
As Smith points out, the 4th source is the most challenging, as it is less salient and in the uncertain future. Smith suggests this is where most people need help.
In the case of pet ownership, the challenge is appropriately evaluating the criterion well into the future. The questions will help. Also, a financial perspective going beyond the here-and-now expense outlay is critical. This is why the long-term retirement impact is evaluated. The idea is not to either encourage or discourage pet ownership. The idea is to provide a decision process so you have conviction in your confidence about the pet ownership decision.
All 4 sources of moral approval must be considered and trade-offs considered. In general, the first 3 sources have the ability to say “no” if the interaction doesn’t meet their needs.
For source 1, some pet sellers evaluate the fitness of the buyer. Also, a pet seller may use price as a way to discourage less committed buyers.
For source 2, the pet buyer will confirm they can afford the pet and it meets many of the criteria mentioned earlier.
For source 3, local rules and customs will impact the pet decision. An apartment building will have rules about pets. Local laws may also impact the decision. For example, pets may not be allowed in certain public places and there are generally rules about cleaning up pet waste.
However, the 4th source does not have as much ability to say “no.” For example, the future you as a pet owner in ten years is challenging to understand. Long-term financial impacts are knowable but may not be properly weighted in the decision.
[ix] Hulett, The subtleties of lending discrimination, The Curiosity Vine, 2022 - see section 1 for the generalized economic discrimination framework.
[xi] In the world of statistics and linear regression modeling, this correlation between independent criteria is known as "multicollinearity." Data scientists are aware of the havoc collinear independent variables create. Interestingly, the answer for resolving collinearity, whether in an AI or HI-based model, is essentially the same. In statistics, the AI will perform feature engineering that separates the information value of the correlated independent variables to create new, uncorrelated variables. This way, the information is maintained in a way that leads to the best-modeled predictor. In the case of our HI personal model, you will do this in a more intuitive manner, but the outcome is the same - a better predictor.