AI and human psychology. 

I did my undergraduate in Psychology and went into to advertising at WPP 2018 from there I was fascinated in “the algorithm”. 



What is “the algorithm”. 

In marketing, when people refer to "the algorithm," they are typically referring to the mathematical formulas and computational processes that govern how online platforms, such as search engines and social media sites, determine which content to show to users. These algorithms are designed to analyze various factors, such as user behavior, preferences, and relevance, in order to deliver personalized and targeted content to individuals. 

Facebook, for example, began using algorithms to personalize users' news feeds in 2006. Initially, the news feed displayed content in a chronological order. However, as the platform grew and the amount of content increased, Facebook implemented algorithms to prioritize and customize the content shown to users based on their preferences, interactions, and relevancy.

Twitter introduced its algorithmic timeline in 2016. Prior to that, Twitter displayed tweets in a strictly chronological order. However, with the implementation of the algorithm, Twitter started curating users' timelines to show the most relevant and engaging tweets first, based on factors like user interactions and content quality.

Where is AI involved? 

Over time, as the amount of content on Facebook increased and user preferences became more diverse, the platform began incorporating AI techniques to curate and personalize the news feed. AI algorithms were developed to analyze user behavior, engagement patterns, and content relevance to prioritize and display the most relevant posts for individual users. 

In 2013, Facebook introduced a major update to its news feed algorithm, placing a greater emphasis on showing higher-quality content and reducing the visibility of low-quality or spammy posts. This update aimed to improve the overall user experience and increase the relevance of the content displayed.

Throughout this period, Facebook progressively integrated AI techniques to analyze user behavior, engagement patterns, and content relevance. By leveraging machine learning and data analysis, Facebook aimed to better understand users' preferences and deliver a more personalized news feed experience.

When posts are shown based solely on high-ranking criteria, the focus is typically on popularity metrics such as engagement (likes, comments, shares) or recency. This approach can result in a feed or search results that prioritize content based on broad popularity or recentness. It may not take into account the specific interests, preferences, or relevance to individual users.

In contrast, when AI algorithms are employed, they analyze a variety of factors beyond popularity and recency. These algorithms consider user-specific data, such as browsing history, past interactions, demographic information, and interests, to deliver more personalized and relevant content. AI algorithms aim to predict what content an individual user is most likely to engage with and enjoy based on their unique profile.

AI is trained is similar to the human brain.

In perceptual psychology, the human brain engages in a complex process of understanding and recognising objects, such as chairs, based on visual information. Similarly, in training an AI model, a similar process occurs, albeit in a computational manner.

When training an AI model to recognise chairs, thousands of pictures of chairs from various angles and types are presented to the model. During this training process, the AI model learns to identify common patterns and features that are indicative of a chair's visual representation.



The AI model analyses the images by examining the individual pixels, identifying edges, shapes, textures, and other visual cues that are characteristic of chairs. Through a series of computations, the model develops a representation or "essence" of a chair that encapsulates the shared features found across the training images.

 

This representation can be thought of as a learned set of weights or parameters within the AI model that capture the statistical relationships between the pixels in the images. These weights enable the model to classify new images as chairs or non-chairs based on the similarity of their pixel patterns to the learned representation. The essence of a chair within the AI model is abstract because it is not a direct replica of any specific chair image or a precise template of what a chair should look like. Instead, it represents a statistical generalisation of the visual patterns shared across the training images.

Capta Images have for a long time been used to train data sets on obscure category items.

Similarly, in perceptual psychology, the human brain processes visual information and identifies shared features and patterns across different chair stimuli. Through experience and exposure to a variety of chairs, the brain develops an understanding of what makes an object a chair. This process involves identifying key visual features, such as shape, size, color, and texture, that are indicative of a chair.

Both in AI training and perceptual psychology, the essence or representation of a chair is derived from analysing and extracting meaningful visual information. While the AI model's approach is based on computational calculations, the human brain's process is driven by cognitive mechanisms and neural processing. The human brain consists of approximately 86 billion neurons, and each neuron can generate electrical impulses known as action potentials or spikes.

(Principles of Neural Science" by Eric R. Kandel, James H. Schwartz, Thomas M. Jessell, Steven A. Siegelbaum, and A. J. Hudspeth)

Let’s consider this chair example and how it compares to finding you content.

So… the AI model in the last example is looking at all the pixels of various chairs, has encoded in it ‘the essence’ of a chair,  and then, if given an image, is seeing if it can accurately predict if the image is “a chair.”

Just as the AI model analyses pixels to identify the essence of a chair, platforms leverage the diverse data points collected from users to create a profile or representation of individual preferences and characteristics. These data points serve as indicators that help platforms predict and recommend content that users are likely to engage with.


For instance, if multiple users who share similar characteristics, such as watching three specific videos until the end, also watch a fourth video to completion, the platform can use this collective behaviour to predict that individuals with similar traits may also find the fourth video appealing. The platform then utilises this comparison of data points to personalise content recommendations and tailor the user experience.

In both examples, the AI model and the platform's personalisation algorithms rely on analysing data points and patterns. The AI model decodes the essence of a chair from pixel data, while the platform deciphers user preferences and behaviours from various data points. By finding similarities between users with comparable characteristics, the platforms can make predictions about user preferences and optimise content delivery.

Conclusions.

In conclusion, the field of machine learning and AI intersects with human psychology in various ways. The algorithms used in marketing, often referred to as "the algorithm," employ computational processes to personalize and curate content based on user behavior and preferences. Over time, AI techniques have been integrated to analyse user engagement patterns and deliver more tailored content.


AI and machine learning models, like GLMs, share similarities with human psychology in terms of learning from experience, pattern recognition, adaptation to new situations, feedback-driven learning, generalization, complex decision-making, incremental learning, and flexibility. While AI models are not equivalent to human cognition, they can process multiple variables and exhibit some level of contextual understanding.

Training AI models to recognize objects, such as chairs, involves presenting large datasets and enabling the models to identify common visual patterns. This process is akin to how the human brain perceives and understands objects through complex cognitive mechanisms. Both AI models and the human brain analyze visual information, identify shared features, and develop an abstract representation or essence.

While AI models provide statistical generalizations based on learned representations, the human brain's perceptual processes involve neural activity and cognitive mechanisms. The human brain's perception of objects encompasses a broader contextual, semantic, and conceptual understanding beyond statistical patterns.

Understanding the similarities and differences between AI and human psychology can provide insights into the capabilities of AI systems while appreciating the unique complexities of your own experience and perception.

Next
Next

Notes: Exploring the death of ‘the aura’