Why People Share: The Psychology Behind "Going Viral" & Catching Unicorns with GLTR

Why People Share: The Psychology Behind "Going Viral" & Catching Unicorns with GLTR

Introduction:

In today's digital age, the concept of going viral has become a sought-after phenomenon. Whether it's a social media post, a news article, or a video, everyone wants their content to be widely shared and recognized. But what motivates people to share content? And how can we detect automatically generated text? In this article, we will explore the psychology behind viral sharing and introduce GLTR, a tool for detecting automatically generated text.

The Psychology Behind Viral Sharing:

1. Status: One of the primary motivations for sharing is the desire for status. As pack animals, humans are constantly thinking about their perceived status and where they fit in. By sharing content associated with high-status individuals or exclusive products, people try to elevate their own status within their social circles.

2. Identity Projection: Sharing content allows individuals to showcase their identity and have their point of view validated. People seek validation in their beliefs and opinions, and sharing content that aligns with their identity helps them feel vindicated.

3. Safety: Fear is deeply ingrained in our psychology, and if we sense danger, our brains instinctively pay attention. Sharing content that provides information about potential threats to safety, such as neighborhood crime maps, helps people feel more secure and protected.

4. Order: Some individuals are highly motivated to share tools and resources that help them optimize and organize their lives. Sharing such content not only makes others perceive them as more organized and efficient but also helps them bring others into the same protocol of organization.

5. Novelty: Humans are naturally drawn to novelty. We are constantly seeking the next new thing, and sharing new products or information makes us appear ahead of the curve. Sharing content related to novelty boosts our self-image and makes us feel good about ourselves.

6. Validation: A significant aspect of sharing is the desire for validation. People want to feel positive about themselves and their place in the world. By sharing their achievements or results, individuals seek validation and a boost to their self-esteem.

7. Voyeurism: People also share content that allows others to experience things vicariously. Whether it's sharing moments of joy or indulging in schadenfreude, the enjoyment derived from sharing vicarious experiences motivates individuals to share.

Detecting Automatically Generated Text with GLTR:

GLTR (Getting to the Right) is a tool designed to inspect the visual footprint of automatically generated text. It helps in the forensic analysis of whether a text has been generated by an automatic system or a human. Traditional language models generate text that closely resembles what a human would have picked in a similar context. This similarity arises from accurate distributional estimates of word probabilities in a given context.

To prevent the generation of text that mimics human choice, forensic techniques like GLTR are necessary. GLTR leverages the same language models used for generating fake text to detect automatically generated content. By ranking all the words the model knows, GLTR analyzes how likely the observed following word ranks.

The tool provides a display where hovering over a word presents the top 5 predicted words, their associated probabilities, and the position of the following word. This insight allows users to explore what a model would have predicted. Additionally, GLTR shows histograms that aggregate information over the entire text, indicating the model's probability assignments and level of surprise.

Although GLTR has its limitations, such as the inability to detect large-scale abuse and the requirement of advanced language knowledge to evaluate uncommon words, it can spark the development of similar ideas that work at a greater scale. Adversarial sampling schemes may result in worse text generation as models are forced to generate unlikely words.

Conclusion:

Understanding the psychology behind viral sharing can help individuals and businesses create content that resonates with their target audience. By tapping into motivations like status, identity projection, safety, order, novelty, validation, and voyeurism, content creators can increase the likelihood of their content going viral.

On the other hand, tools like GLTR enable the detection of automatically generated text, protecting against the spread of misinformation and fake content. While GLTR has its limitations, it opens the door for further advancements in the field of detecting and preventing the circulation of automatically generated text.

Actionable Advice:

1. Understand Your Audience: Identify the motivations that drive your target audience to share content. Tailor your content to align with those motivations, increasing the chances of it being shared.

2. Reduce Friction to Sharing: Design your product or content in a way that minimizes the effort and thinking required to share. Make sharing seamless and easy for your audience.

3. Use GLTR as a Forensic Tool: Utilize GLTR or similar tools to inspect the visual footprint of automatically generated text. Stay vigilant against the spread of fake content and misinformation.

By combining an understanding of the psychology behind viral sharing and leveraging tools like GLTR, individuals and businesses can navigate the digital landscape more effectively and create content that resonates with their audience.