Delve's AI assist you in your qualitative research, but it won't replace your role as a critical thinker in the research process.
After you've created codes and added descriptions, Delve's feature, "Apply Codes Using AI" can apply your existing codes to your transcript.
AI will help apply your codes, but they may not be accurate in interpreting the meaning behind your data. You, as a humen, are still required to draw insightful conclusions and themes from your data.
Only available for AI beta program participants.
To join Delve's beta program, accept the Beta Agreement that appears on the AI Chat page.
By leveraging your code names and code descriptions, the AI Assistant reads your transcript, identifying and coding snippets that match the criteria set by your codes.
This process is known as deductive coding, a method characterized by the use of an established codebook to systematically apply codes to the transcript. Since the AI Assistant utilizes your existing codebook for this purpose, it employs a deductive coding approach.
Watch how AI codes your transcript using existing codes:
Before You Start
Have Transcripts Uploaded: The AI feature will code existing transcripts
Add Code Names and Code Descriptions: The AI uses your code names and the first paragraph of your code descriptions to decide how it codes.
Steps to Apply Codes Using AI in Your Transcript
Select Your Transcript: Navigate to and click on the transcript you wish to code with AI.
Access Transcript Options: Locate and click the "..." icon positioned in the upper right corner of the screen, and select "Apply Codes Using AI"
Start Coding: Click on the blue "Apply codes using AI" button to start the coding process.
Review Coded Snippets: Once the AI completes its task, coded snippets will be displayed in a list within a pop-up modal.
If you don't see snippets appear after a few minutes, try refreshing your page.
Important Note: Feel free to navigate away from the window at any point. Your departure won't disrupt the coding process. You can return at your convenience to check on the progress or review the completed work.
See explanations for AI's coding decisions in memos
Once the AI finished applying codes, you'll be able to see the explanation for how the codes were applied within the memos in the snippet.
Ask AI follow up questions in memos.
If you have questions about the AI's coding choices, you can ask follow up questions in the memos by tagging @ai_assistant.
Enhancing AI Coding Through Iteration
The effectiveness of AI coding is directly linked to the clarity and specificity of your code names and descriptions. Much like any qualitative research project, we recommend adjusting your codebook iteratively until the AI-generated codes meet your expectations. With AI, this iterative process is accelerated, allowing for faster improvements.
If the initial coding results fall short of expectations, don't be discouraged. Undo the AI's coding for the transcript and embark on a cycle of iteration and refinement. Adjust code names and descriptions based on the shortcomings identified, and rerun the process. This iterative approach enables continuous enhancement of coding precision, ensuring that your analysis evolves to meet the demands of your data effectively.
Watch how to remove codes applied using AI
Steps for Removing Codes Using AI
Click the transcript you want to code using AI
Click on the "..." icon in the upper right hand corner of the screen
Click the option "Apply Codes Using AI"
Click on the button, "Remove codes applied with AI"
This will remove all the coding done by AI on this specific transcript. You can now have the AI recode your transcript.
Strategies for improving your codebook
Creating a comprehensive and effective codebook is crucial for enhancing the accuracy and efficiency of AI-driven coding processes. Here are refined strategies to improve your codebook, ensuring it serves as a robust foundation for AI to analyze and categorize data:
(Note: Many of these strategies are best practices even when you are not using AI!)
1. Incorporate Detailed Code Descriptions
Why: Detailed descriptions provide clear guidelines on how each code should be applied, reducing ambiguity and improving the AI's ability to match text snippets accurately.
How: For each code, write a description that covers its scope and intended use.
2. Refine Code Descriptions for Clarity
Why: Clean and concise descriptions help the AI to understand the context and apply codes more accurately.
How: Regularly review and edit your code descriptions to remove redundant words, clarify meaning, and ensure consistency across the codebook.
3. Ensure Codes Are Distinct and Non-Overlapping
Why: Overlapping or vague distinctions between codes can lead to incorrect or inconsistent coding by AI.
How: Evaluate your codes for overlap and merge or redefine codes as necessary to ensure each has a unique and clear purpose.
4. Introduce New Codes to Broaden Your Codebook for Better Coverage
Why: A broader codebook prevents AI from overusing other codes by providing more specific options, ensuring every snippet is appropriately categorized. This approach enhances the depth of analysis, as AI utilizes the detailed structure you provide to organize data more accurately.
How: Regularly review your data for uncovered themes or patterns, adding new codes to address these gaps. This keeps your codebook relevant and comprehensive, improving data organization and analysis.
5. Implement an 'Other' or 'Miscellaneous' Code
Why: There will always be data that doesn't neatly fit into your predefined categories. An 'Other' category helps identify these outliers.
How: Create a code for data that doesn't match existing codes, and periodically review these instances to determine if a new code is warranted.
6. Review and Update the Codebook
Why: As your project evolves, so too will the data you're analyzing. Keeping your codebook up-to-date ensures it remains relevant and effective.
How: Schedule regular reviews of your codebook to adjust codes, refine descriptions, and incorporate new insights from ongoing analysis.
7. Engage with Your Data
Why: Direct engagement with your data can reveal nuances that might not be immediately apparent, guiding more informed updates to your codebook.
How: Spend time coding manually, especially in the early stages of project development, to gain insights that can refine AI coding strategies.
8. Leverage AI Memos
Why: AI offers invaluable insights into coding challenges through detailed memos explaining its decisions.
How: If you disagree with the AI's coding, review it's memo to understand it's reasoning. You can use this information to make adjustments to your codebook.
By following these strategies, you can build a more effective and AI-compatible codebook, leading to more accurate and insightful data analysis.
Frequently Asked Questions
Q: How many transcripts can AI research assistant code at a time.
The AI research assistant can code one transcript at a time.
Q: How do I get access to the AI assistant?
The AI Assistant is currently available as part of our open beta program. If you are interested in joining the beta program, please accept the terms on the AI Chat page.
Want to learn end to end qualitative coding with AI?
Watch our 30 minute webinar where we break down the concepts of qualitative coding and how AI can be used to assist your process.