Systematically speaking, the key to identifying challenges lies in examining fundamental issues, not just their consequences. For example, if we want to fix the flaws in the learning process, we must first redefine the roots of deep learning and its underlying mechanics. We may even need to redefine them repeatedly to understand how to solve the problems arising from mind-based technologies.

Let me explain what I mean through one of the most debated topics of our time: the mental laziness caused by the way AI is rewriting our brain’s habits. To understand this, we need to look at the dynamics of deep learning in the brain. By grasping this process through interdisciplinary research, we might find ways to make AI learning feel more like natural deep learning.

The goal isn’t just to know the biochemistry of cells. Before looking at what happens inside an organism, we should ask:

Why do we usually prefer learning through AI over the effortful, traditional human way?

You might say the answer is obvious: because learning with technology is effortless and fast.

As a learning specialist, I’d like to answer this from a theoretical perspective.

 First, we must accept a reality: Human deep learning is naturally a challenging process. It is fundamentally different from the vast amounts of data we consume today through formal or informal education assisted by LLMs.

The Logic of Immediate Reward: From Skinner to the Present

There is strong research showing that learners prefer a small, immediate reward over a larger, delayed one. This was first highlighted by B.F. Skinner (1953), the pioneer of operant conditioning. (I’ve previously written about how this connects to AI. )

Later, others expanded on this effortless reward preference. In short, according to the behavioral economics of Skinner’s theory, humans look for shortcuts. 

AI is currently the ultimate shortcut, giving the best answer in seconds without any real struggle. From this view, it’s not just about the mind; it’s about behavioral economics.

A behavior that leads to a quick reward will always be repeated.

Richard Herrnstein (1961), a student of Skinner’s, developed a mathematical formula called the Matching Law. He showed that organisms don’t just look at one reward; they choose between options. If given two choices, a living being will put its energy into the one that pays off faster and more directly. 

In behavioral economics, this phenomenon is known as temporal discounting (Ainslie, 1975). The value of a reward drops the longer you have to wait for it. Simply put, the reward loses its shine in the organism’s mind because it requires patience.

We observe this phenomenon every day with AI users, particularly those utilizing ChatGPT. Students, for instance, might feel that spending hours writing a thesis is stupid or inefficient when they can get an answer in a split second. They don’t just feel productive; they feel smart for bypassing the effort. 

Even if you tell them that the struggle is what actually builds their brain, they often won’t listen. They choose the immediate payout over the long-term value. 

Evolutionary psychology explains this too: an immediate reward is guaranteed, while a future one is uncertain. Since we are wired for survival, we grab what’s available now.

Brain Biochemistry and the Deep Learning Process

When we learn something deeply, three key things happen at a neurological level:

1. Exposure to New Information: The nervous system makes its first contact with data for which it has no existing pattern.

2. Cognitive Load: This is that stuck feeling when a mental process is harder than expected. It’s the effort the brain needs to process unfamiliar data (Sweller, 1988). This friction is essential.

3. Processing and Protein Synthesis: If the information is processed correctly, chemical signals trigger the creation of proteins that physically change the brain’s structure to store that knowledge (Kandel, 2001).

This is why sleep is so vital. Most of this protein synthesis happens while we rest. 

One of the most beautiful parts of learning is when we stop thinking about a problem, but our brain keeps working on it. 

Through the Default Mode Network or DMN (Raichle, 2015), the brain makes random, creative connections. This is where true creativity is born.

Toward Friction-Based AI

If deep learning is the result of protein synthesis triggered by challenge, then the paradox of modern AI is clear: By removing the friction, technology is removing the learning. 

We are facing a biological crisis where human brains, instead of producing genius and problem-solving skills, are becoming mere terminals for receiving quick hits of dopamine.

My proposal is simple: How can we turn AI from a passive answer-giver into a Cognitive Challenging Provocateur? 

We need to design models that don’t bypass cognitive load but manage it in a personalized way. 

I call this Friction-based AI; a model where algorithms are programmed not for the shortest path, but for the most effective learning path. This is an open invitation to researchers, neuroscientists, and AI architects to collaborate on this new paradigm. My ideas are ready to be turned into actionable proposals.

As a final note, I believe the way we interact with AI is a skill in itself. Even if everyone has the same tools, the results aren’t equal. Efficiency depends on the how. 

I am currently developing a startup idea to address these exact challenges in EdTech.It’s EdTechxDr. Atefeh F.

References

• Ainslie, G. (1975). Specious reward: A behavioral theory of impulsiveness and impulse control. Psychological Bulletin.

• Herrnstein, R. J. (1961). Relative prevalence of response in relation to the relative frequency of reinforcement. Journal of the Experimental Analysis of Behavior.

• Kandel, E. R. (2001). The Molecular Biology of Memory Storage: A Dialogue Between Genes and Synapses. Science.

• Raichle, M. E. (2015). The Brain’s Default Mode Network. Annual Review of Neuroscience.

• Skinner, B. F. (1953). Science and Human Behavior. Simon and Schuster.

• Sweller, J. (1988). Cognitive Load During Problem Solving: Effects on Learning. Cognitive Science.

The post Why Biological Learning Demands the Friction We Seek to Delete? appeared first on Medika Life.

Generally, artificial intelligence is used to emulate human behavior and serve humanity (which seems to be the case). In that case, it will inevitably have to study all human sciences as sources for understanding human nature and essence.

As has been said many times, theories are powerful resources that generate new research and hypotheses. Sometimes, they also discard previously confirmed hypotheses that lack the necessary efficacy in the new era. This flexibility enables adaptation and changes required in an era of speed and modernity. Therefore, theories provide us more flexibility, predictability, and a life with greater peace of mind. 

In this case, it can be said that the possibility of creating a Happy Modernity in an era of confusion caused by the instant speed of artificial intelligence technology will not be out of reach. 

As mentioned, theories related to human sciences, including social sciences, psychology, and behavioral sciences, can be the flag bearers of this change and the construction of a better world.

So far, much has been said about cognitive sciences and neuroscience. Among these, behavioral studies and behaviorist theories have received less attention. This article discusses the importance of the behaviorist approach, particularly the conditioning of Skinner and its interaction with artificial intelligence, albeit very briefly and generally.

About B.F. Skinner and Operant Conditioning

B.F. Skinner, the renowned American psychologist born in 1904, revolutionized the field of behavioral psychology with his experimental studies on operant conditioning.

 His experiments with rats and pigeons demonstrated how behavior could be shaped through reinforcement and subsequent consequences, laying the foundations for modern behaviorism. 

See this link about his fame experiment :

During the 1930s, B. F. Skinner proposed the theory of operant conditioning, which states that behavior change and learning occur as the outcomes or effects of punishment and reinforcement.

Skinner’s influence extended beyond psychology and impacted fields such as education, technology, and even artificial intelligence algorithms. His theory inspired the development of artificial intelligence algorithms, particularly in reinforcement learning, where agents learn to optimize behavior based on rewards and punishments, reflecting Skinner’s principles. 

 If we were to discuss Skinner’s entire theory and its inspiring effects on the scientific world, we would have to dedicate several articles to this topic.  Therefore, the main focus of this article is to explore the role of this important psychological theory on algorithms and the AI age.

In this case, the essence of Skinner‘s theory can be summarized as the impact of behavioral consequences on the shaping and continuing behavior or responses. 

This simple principle, which is the most important result of Skinner’s experiments and the essence of his theory of operant conditioning, has alone inspired fundamental developments in areas such as programmed learning and teaching machines, distance education, behavior modification, psychotherapy or behavior therapy, medicine and neurofeedback, principles of child-rearing, and currently artificial intelligence and machine learning. 

However, as usual, it should be noted that this important psychological theory needs to be better understood, and after recognizing its flaws and criticisms, its benefits and principles should be taken into account more in building the world of artificial intelligence and applying behavioral principles in designing artificial intelligence tools. 

 Therefore, by considering what critics of Skinner’s theory say, that it is too mechanical and radical and downplays the role of cognitive factors and human existence, we can take advantage of its benefits and key points, such as the crucial effect of consequences on behavior and response, as an essential key to designing better technology and taking steps towards a” Happy Modernity.”

Similarities of the Response Consequence Effect in Skinner’s Theory and AI Algorithms

Please consider the following points if you want a simple yet practical comparison. Then, you’ll know that understanding this comparison can help us better lead advanced artificial intelligence machines, regardless of the criticisms against Skinnerian behaviorism.

 Indeed, as one of the most influential contemporary psychologists, Skinner’s dream was precisely this: to create a disciplined behavioral technology and engineering that would enhance life and make it easier! 

Please consider these fundamentals:” Reinforcement” (both positive and negative) influences the repetition and likelihood of responses in organisms. “Positive reinforcement” increases the probability of behavior by its presence, while “negative reinforcement” increases the likelihood of response by its removal. However, the goal remains clear: the “consequence “influences behavior!

• Both in Skinnerian theory and in artificial intelligence algorithms, positive reinforcement is the same as reward, and negative reinforcement includes punishment and penalty.
• Another common aspect between Skinner’s operant conditioning and artificial intelligence is learning through interaction with the environment!  Most organisms learn through interaction and by gaining experience in the surrounding world.
• In operant conditioning and artificial intelligence, a relatively straightforward cycle is repeated: action, observation, and feedback.

This cycle is repeated until the desired outcomes are achieved! In addition to the points mentioned, operant conditioning has been directly incorporated into the design of reinforcement learning algorithms. Techniques such as Q-learning are model-free, value-based, off-policy algorithms that find the best series of actions based on the agent’s current state.

The term “Q” stands for quality, representing how valuable the action is in maximizing future rewards. The applications of this symbiosis between operant conditioning and reinforcement learning are extensive and diverse.

I have some suggestions for the useful Application of Skinner‘s Theory in Artificial Intelligence Technology.

Here, I have briefly listed more applications of operant conditioning theory in artificial intelligence technologies. Furthermore, I am very eager to hear your ideas and suggestions after reading these insights and my ideas.

Applications of Operant Conditioning in Artificial Intelligence: Bridging Behaviorism and Technology

From what was discussed in the previous section of this article, the applications of operant conditioning in artificial intelligence are almost evident.  However, if we want to define this synergy more specifically, my suggestions are as follows:

• In robotics, artificial intelligence tools can perform complex tasks through reinforcement learning, such as navigating unfamiliar environments or manipulating objects precisely.

• In the realm of autonomous vehicles, it appears that reinforcement learning mechanisms based on operant conditioning enable continuous adaptation to road conditions and traffic patterns. Thus, employing the simple principle of consequences on response leads to increased road safety and security by autonomous vehicles. 

• Besides robotics and autonomous intelligent systems, reinforcement learning has applications in various domains such as finance, healthcare, and gaming.

Notably, in designing principles of behavior therapy and therapeutic interventions, using the principle of response consequence and feedback is considered one of the influential principles in treating behavioral disorders.

Especially in medicine and clinical psychology, where discussing diagnosis and treatment through artificial intelligence is very hot, applying behavior therapy based on operant conditioning is inevitable.

Applying these principles in neurofeedback is highly recommended and has been the subject of extensive research for years.  In the world of education and learning through artificial intelligence algorithms, one of the primary principles of artificial intelligence application in education is personalized and learner-based learning.

It is implicit that this key principle of individual learning based on personal speed and rapid feedback is rooted in the same core principle of Skinner’s theory, which is the individual learning system based on response consequences.

Artificial intelligence in schools and higher education in advanced and developed countries is rapidly developing, and its most important feature is personalized learning based on consequences.  These consequences or feedback are provided to students by their learning partner and mentor, which is artificial intelligence. 

Another application is RLHF, which means “Reinforcement Learning with Human Feedback.” It’s a new area where computers learn from regular signals and direct input from people. This mix helps AI systems improve at tasks like making recommendations or controlling robots. RLHF is exciting because it lets humans and machines work together, making AI systems smarter and easier to understand. See this link https://johnnosta.medium.com/insights-on-ai-understanding-rlhf-f4b79cfcbdc8

 In general, artificial intelligence promises a revolutionary breakthrough in various fields through reinforcement learning and behavior optimization, from education and optimization of financial strategies to personalization of psychological and medical treatments.

However, significant ethical considerations are also required in this remarkable historical leap. As artificial intelligence systems increasingly become capable of shaping human behavior and guiding individual and social life, autonomy, privacy, and accountability issues take center stage. 

Therefore, ensuring that ethical principles and human values guide the application of reinforcement learning in artificial intelligence is essential to protect against unintended consequences and harmful outcomes.

In conclusion, B.F. Skinner’s operant conditioning theory has significantly shaped the landscape of artificial intelligence algorithms, particularly in reinforcement learning.

 By grasping the essence of behavior modification and the profound impact of consequences on behavior, AI systems stand to benefit across diverse fields, from robotics to healthcare and education.

However, it’s imperative to remain cognizant of ethical considerations, ensuring that AI deployment aligns with human values and ethical principles to mitigate potential risks and amplify societal benefits.

I invite you to read my articles on applications of behavioral theories in AI algorithms, available on MedikaLife and LinkedIn, for a deeper dive into this fascinating intersection of psychology and technology and to get “Happy Modernity” in the AI era.

The post From Skinner’s Operant Conditioning to Artificial Intelligence’s Algorithms appeared first on Medika Life.