Lately, I’ve seen a lot of people concerned about ‘bias’ in AI, how they affect their models and so on. It has gotten to the point that I can no longer ignore the need for myself to formulate a straight way of thinking, to hopefully; someday recreate what a good and ethical AI system should be; in harmony with the natural state of the universe.
The reason for me to write this not so technical and lack of more in-depth knowledge article is due to the so many approaches, wanderings, and questionings on diverse blogs, internet posts, and articles were written about it. I have noticed very well illustrated examples with very complex terminology that I can not fully understand; but it seems to be relevant to the work, study, and research that they are doing.
I write this article not to disrespect anyone else’s hard work, but to express my ideas and way of thinking; and how from my perspective, this area of “expertise” needs an extra study to fully maximize the desired results using AI.
A little bit about Statistical bias
For centuries, great mathematicians have tried and succeeded to explain how our surrounding universe works in terms of probability. We try so hard to predict the outcome on many given actions; the unknown as uncertainty has always pushed us to greater mathematical lengths to have an advantage and be in a better position after the action is performed.
Bias a historically bad actor?
Historically, the word “bias” has always been internalized as the “bad” actor since it helps someone “cheat” the system and take advantage to win diverse actions most likely when bets and money are involved.
If it is true, and from my perspective; this way of projecting results on immediate actions based on multiple historical or modeling data will most of the time work. But also, this kind of model has been based and envisioned by the “gambling” perspective. That is… “if I do this, I will win that” and here is where I tend to deviate a little bit from what I call ‘classical bias’ versus what I call ‘modern cognitive bias’ directed towards AI.
How the mind works -my perspective-
There are many studies, results and cognitive theories about “how the mind works”. This a field in which I am not an expert. Yet, I am courageous enough to describe how our mind works in my own words, so bear with me.
When you google the term “How the mind works”, we get a series of results and best seller books from renowned researches and scientists; yet, all come to the same biological evolutionary conclusions a good example can be seen here: Pinker, S. (2009). How the Mind Works (1997/2009) (2009th ed.). New York, NY: W. W. Norton & Company.
Don’t get me wrong, I believe that these kind of results are precious and makes us advance our understanding; but we shouldn’t be taking this as our “plateau” with nothing else above. And here is where I come in…
Above our current “plateau”
My current way of thinking is beyond natural or biological evolution; in fact, it goes back to the big bang when the chaos generated by it distributed atoms all over our existing universe. I will explain why.
Not too long ago, I read this phenomenal book: Decoding the universe: how the new science of information is explaining everything in the cosmos, from our brains to black holes by Seife, Charles. The author and his particular way of storytelling changed my way of thinking, and my way of seeing things in a completely different and unexpected way… and this is the base for my next analysis.
BTW is highly recommended.
Entropy
Is interesting that in this book, I actually conceptualized and internalized the concept of Entropy in which Claude Shannon (father of information theory) came up with a formula that already existed in Thermodynamics and thanks to John von Neumann (the last representative of the great mathematicians) name it Entropy overall. And from here moving forward, my mind started to connect dots in a completely different way as I used to.
My point is… thanks to Entropy, I started thinking about the Universe and the black holes; atoms and speed of light; quantum state and our brain; and here is where I am writing a completely different, perhaps not so complex and possible bizarre idea.
The Brain in Quantum State?
As I was mentioning before, our bodies and in particular our brains are made of atoms, and atoms are the ones that form us as individuals with physical differences and cognitive differences. Hence, I tend to be inclined to the way of thinking that our brains work in the same way as the quantum mechanics describe. That means that our brain atoms are superposed in a quantum state in a physical world (our body) and in a non-physical world (where ideas, intuition, and decisions) are made. I make that distinction to express those ideas, intuition and decisions are intangible.
Classical Bias modeling Quantum State?
Now, from my perspective; if we take the above way of thinking and we extrapolate it into modern machine learning or even AI algorithms and models; those bias that we want to avoid will not satisfy such governing “non-physical” rules by inferencing with “physical” rules. Something like the theory of relativity and quantum theory; yet the two are the biggest discovery compared to our well-known story when an apple hit Newton’s head. The two are complements of each other. And here is where I believe the current need to expand our current knowledge to satisfy the current need in which we are so desperately trying to solve ‘cognitive bias’.
What about Neuroscience?
If it is true, Wikipedia describes Neuroscience (or neurobiology) is the scientific study of the nervous system. It is also a multidisciplinary branch of biology and combines physiology, anatomy, molecular biology, developmental biology, cytology, mathematical modeling, and psychology to understand the fundamental and emergent properties of neurons and neural circuits. The understanding of the biological basis of learning, memory, behavior, perception, and consciousness has been described by Eric Kandel as the “ultimate challenge” of the biological sciences.
I believe that this area is creating tremendous breakthroughs and need to be evaluated in terms of perhaps creating a bridge from the physical world to the non-so-physical for the quantum state that I’ve been formulating above; perhaps this is just one piece of the puzzle.
I’m also thinking about the theory of chaos (fractals)…
I’ve exposed in what I hope simplistic and short terms my way of thinking; but also for some reason I am thinking of the beauty that chaos can bring. I don’t mean this to be in bad terms, but chaos in a mathematical world such as the fractals in which they describe infinitely complicated abstract objects used to describe and simulate naturally occurring processes. and for this, I am “assuming” that cognitive bias is a natural occurrence unless someone refutes it in a very expedited notion that makes sense.
Yet again, “assuming” in the classical statistical world might not be seeing in a good way; but my way of thinking is beyond that, hence I could infer that ‘cognitive bias’ is a chaotic natural occurring phenomena following the same properties as the fractals do; I might be wrong, but from my perspective seems to comply with the governing rules.
What I am trying to explain is that if we limit our ‘cognitive bias’ with ‘classical bias’ limitations for small portions of our model or even our study, we are limiting nature itself keep the respective balance of the balanced natural chaotic world by creating an “unbalanced one”. Perhaps, we are currently being myopic and since we are not able to see a “full-fledged cognitive bias fractal” from our data or given models, we take shortcuts on what is known to us, limiting our maximum return potential without even knowing.
but what about our DNA?
Something interesting nowadays is that we are aware of our physical composition; we know that we are being transmitted chains of information in our DNA from the moment we are conceived, and this information is passed for generations.
I bring this up for me to further explain from my perspective that ‘cognitive bias’ is a naturally occurring phenomenon since in many ways we are born with some of them, others we learn them down the road from experience and culture.
Bias based on intuition?
Another idea is that perhaps we could train our AI model to create some sort of artificial intuition, but yet again we get to the similarly closed door in which intuition can not be fully dictated since is an intangible attribute to human beings and not easily transposed to any machine.
Conclusion
From the above, we could conclude that we need further studies to understand how the mind works in a quantum state. We could extrapolate these ideas and merge our algorithms and models by applying chaotically approaches to cognitive bias. We should not create equal models for different observers since we all have different perceptions. We should be able to create modular AI models that can apply different bias for improved results (I will call these as ‘diversify bias advancements’).
I don’t know if I am correct or if I am wrong. This is just a humble way of exposing ideas for further discussion. Perhaps we could all enhance our knowledge and grow.