Algorithmic Model of Decision Making in the Human Brain(Unfinished)
Abstract
First of all, I want to clarify why I refer to decision-making. As far as I am concerned, an AI model (LLM) currently finds it difficult to process all the information from past interactions. You may notice that when we have a long conversation with Gemini, ChatGPT, or other models, it cannot always capture the details I have mentioned precisely. Sometimes we do not want certain information from the history to be included, but the model still takes it into account. My goal is to make the daily chat function, supported by the LLM in the AI agent, more human-like.
This article aims to learn how the human brain makes decisions in order to better integrate biological knowledge with computer applications — that’s an AI agent’s long-term memory. And seeks to verify whether human decision-making algorithms can be accurately simulated under traditional computer architectures. The last thing, here I do not and can not to verify the correctness of the evidence. So the following content is based on the conclusion of the cited paper and presents a personal ideas.
Main content
division p1
"It is well known that the decision-making process results from communication between the prefrontal cortex (working memory) and hippocampus (long-term memory). However, there are other regions of the brain that play essential roles in making decisions, but their exact mechanisms of action still are unknown. "
That means the current talking message is working memory, and the long term memory is the previous data we input. Although the computer architecture we generally use excels at storing information, it will be a enormous memory storage demand when it comes to video and audio. And please try to recall: is it hard to keep a full and detailed image of a scenario, even if you just watched it thoroughly? Instead, we tend to keep some features of the scenario—its shape, color, types, temperature, and other aspects. Then here is a new question: why and how do we remember these features in our mind? Put more professionally, how does our brain capture, filter, and store this information in our brain?
division p2
"a central tenet of rational decision-making is logical consistency across decisions,regardless of the manner in which available choices are presented. This assumption,known as “'extensionality”or “invariance, is a fundamental axiom of game theory .However, the proposition that human decisions are “description-invariant” is challenged by a wealth of empirical data . Kahneman and Tversky originally described this deviation from rational decision-making,which they termed the “framing effect,” as a key aspect of prospect theory.
Theories of decision-making have tended to emphasize the operation of analytic processes in guiding choice behavior. However, more intuitive or emotional responses can play a key role inhuman decision-making. Thus, when taking decisions under conditions when available information is incomplete or overly com-plex, subjects rely on a number of simplifying heuristics, or efficient rules of thumb, rather than extensive algorithmic processing"
Put it simply, it means that our decisions are not always rational because they can be influenced by emotional and intuitive responses. When we applied it to LLMs, suggests that LLMS don't have human-like emotions or intuition. Instead, they are trained on vast datasets that contain human language, which includes expressions of emotion, rational thought, and intuitive leaps. Their "emotionality" or "intuitive" responses are not based on personal feelings but are learned from patterns in the data they were trained on. Therefore, can we categorize the filtered data that an LLM is trained on as containing elements of rationality (logic, facts), emotionality (sentiment, emotional language), and intuition (heuristics, shortcuts, and biases that are common in human text).
Reference Cites
Saberi Moghadam S, Samsami Khodadad F, Khazaeinezhad V. An Algorithmic Model of Decision Making in the Human Brain. Basic Clin Neurosci. 2019 Sep-Oct;10(5):443-449. doi: 10.32598/bcn.9.10.395. Epub 2019 Sep 1. PMID: 32284833; PMCID: PMC7149951.
Heekeren, H., Marrett, S., Bandettini, P. et al. A general mechanism for perceptual decision-making in the human brain. Nature 431, 859–862 (2004). doi.org/10.1038/nat…
De Martino, B., Kumaran, D., Seymour, B., & Dolan, R. J. (2006). Frames, biases, and rational decision-making in the human brain. Science, 313(5787), 684–687. doi.org/10.1126/sci…
small talk
This is my first time to write a article, and this article may look like a academic style(Since I wanna represent the theoretical ideas behind the long-term memory of ai agent). If there is any copyright issue, please tell me and I will handle it as quickly as I can.
