Research is a search and data relevancy problem. 

Granted, that may not match the popular image of research, which is often portrayed as a brilliant insight followed by a quick trip to Stockholm to claim a Nobel Prize. Nonetheless, much of research consists of: 

• Visiting the library—usually online—and seeking out relevant information, downloading papers, reviewing them, taking notes, and filing them.  

• Engaging in conversations with other researchers, advisors, funders, and so forth. This could be in person, over Zoom, via email and Slack, and so on. 

• Conducting studies and accumulating data.  

• Analyzing data using a variety of tools.  

• Keeping papers and notes about those papers connected. 

• Keeping papers and studies connected.  

• Keeping data connected to studies. 

• Keeping notes of conversations, emails, Slack messages, and so forth, connected to the relevant studies. 

• Writing, reviewing, editing. 

There is a lot of data that needs to stay connected to other data, and that amount only increases over time. Not only do each of the listed steps happen multiple times, they happen over long periods of time as the amount of data grows. Since researchers are often involved in multiple projects, they are going through this data management process frequently, with different papers, different data, different studies. They might need to refer back to any of their studies when writing grant proposals. Studies build on studies, so keeping track of old data is important. Data for publication may need to be anonymized (for example, in social science studies), adding an additional layer of complexity.  

Data Management is Friction 

No matter how phenomenally well-organized you are, repetitive data management tasks are a part of life—one of the less enjoyable parts. It wastes time and energy, especially as the amount of data increases. Search tools, which seem like they should be helpful, are of limited use. Since managing information underpins everything, the more time spent on that task the slower everything goes.  

Applying AI  

The good news is that AI can help with research data management and search. The trick is realizing that you don’t need to train an AI on your data, nor should you—any data an AI is trained on could be inadvertently revealed. Particularly in areas that require strict confidentiality—such as human subjects research—training an AI on confidential research data is a security nightmare. 

Bringing your data to the AI is an approach that is fraught with danger. However, through the use of a little technological jujitsu, you can bring the AI to your data and turn AI into your own personal research assistant. 

Bringing the AI to the Data 

Bringing the AI to the data instead of bringing data to the AI may sound like a play on words, but it represents an innovative approach to taking advantage of the power of AI while minimizing the risks. 

When you bring data to the AI, the data becomes part of the AI’s knowledge base. When you bring the AI to the data, the AI can be used to analyze data and determine relevancy, but it doesn’t remember what it has seen. This second approach requires much less data movement and keeps data safely within your trust boundaries. It also lets you leverage the power of AI to identify relevant data in any location, so you can easily add new databases to the search space.  

Bringing the AI to the data requires AI infrastructure software that provides: 

• Universal connectivity: Data remains in place and the infrastructure software enables the AI to read data for analysis. The AI does not remember the data it sees. 

• AI flexibility: You can use any AI model that is approved by your organization.  

• Advanced metasearch capabilities: Data can be anywhere, including library databases, local or cloud storage, email, Slack, and other collaboration apps. Since data can be anywhere, you need to be able to search everywhere, it needs to be quick, and results need to be reliable and organized by relevancy to your initial query. 

• Prompt enhancement: It’s easy to write a bad AI prompt. Rather than everyone becoming a prompt engineer, the infrastructure software should optimize the prompt to limit the possibilities of the AI returning incorrect information or hallucinating.  

• Ease of use: The infrastructure software should make it easy for everyone to use AI, not just trained experts. 

Fortunately, you don’t have to wait for AI infrastructure software to become available someday. It’s available right now.  

AI Infrastructure Accelerates Research 

SWIRL provides the AI infrastructure software and metasearch capabilities that ends the data scavenger hunt. SWIRL provides an innovative, powerful, and flexible approach to combining AI, search, and data management, turning any AI into your personal research assistant. SWIRL evolves and scales with technological advancements so you can avoid being locked to a specific AI model or vendor.