Project Overview
Project Summary:
The Deep Think project is designed to enhance critical thinking among individuals living
in today’s information-saturated society by enabling users to engage more deeply with the
content they consume.
When users input informational text—such as news articles, blog posts, or URLs—the
system utilizes generative AI to analyze each sentence, breaking it down to identify the
underlying evidence and perspective, while also summarizing the overall content and
inferring the author’s intent. To support this process, the system visually categorizes
information: statements supported by objective data, such as statistics or verified facts, are
marked in green, while those that reflect interpretation or multiple viewpoints are marked in
purple.
Users can click on any sentence to reveal additional context, including supporting links
and alternative perspectives, helping them to understand not only the surface-level meaning,
but also the deeper motivations and implications behind the text. By encouraging users to
consider information from multiple angles and by highlighting the strength and subjectivity
of each claim, Deep Think fosters a habit of critical analysis.
Ultimately, the platform aims to help users form more balanced and well-informed
opinions by promoting thoughtful engagement and reducing bias in the way information is
accepted and understood.
Identifying the Challenge
The Social Problem:
Currently, the development of Generative AI is producing vast amounts of information with
unknown authenticity. Unlike before, much of this information lacks clear sources, contains
factual errors, and is often misleading.
As people become overwhelmed by processing so much content, their critical thinking
weakens, making them more susceptible to biased perspectives. Additionally, fabricated
information and propaganda can easily sway public opinion, leading to misinformation and
division.
This poses a serious threat to social stability, as distorted information can influence
important decisions. If left unaddressed, these issues could lead governments and societies
to make misguided choices based on false or misleading data.
Therefore, it is crucial to develop strategies to verify information, promote media literacy,
and encourage critical thinking to mitigate the negative impact of misinformation in the
digital age.
Innovation and Uniqueness
Why Our Project Stands Out:
"Deep Think" offers significant advantages over traditional web searches and Generative
AI, which often require extensive searching and careful prompt crafting, making them timeconsuming and inefficient. To address this, "Deep Think" streamlines the process by quickly
summarizing content, identifying the author's intent, analyzing key points, and providing
references—all with just a URL or text input.
A key feature is its color-coded system, allowing users to visually distinguish between
factual information and subjective claims. Additionally, by leveraging web crawling and
Generative AI, "Deep Think" enhances critical thinking by presenting multiple perspectives
and relevant sources for each sentence. This helps users engage with information more
effectively, reducing the risk of being misled by biased or false content. Ultimately, "Deep
Think" promotes information literacy and empowers users to make more informed decisions
in today’s digital landscape.
Insights and Development
Learning Journey:
Through this project, we learned the importance of thoroughly understanding API
documentation and the challenges of prompt engineering in generative AI. Designing
effective system prompts required not only technical understanding but also clear logic and
communication skills. We also realized that generative AI isn’t always reliable or suitable for
every task, and integrating it into real-world services requires a flexible and adaptive
approach. These experiences helped us develop a deeper understanding of how to work with
AI in a real development environment.
Development Process:
We encountered several challenges during development. While generative AI services
worked well in the browser, their APIs often had policy limitations. For example, Perplexity’s
API didn’t support web scraping, so we built a custom solution using Python scraping
libraries. As data volumes increased, API response times became a concern, with some
requests taking over 60 seconds. We realized that implementing async processing would be
crucial for improving performance in future projects. We used Jira for task tracking, Notion
for documentation, and GitHub for version control. Following Git Flow and an agile process
helped streamline development and enhance team collaboration. Regular meetings and
code reviews ensured consistency and quick resolution of issues.
Github: https://github.com/Deep-Think-Project
