AI and qualitative research: shifting landscapes

At CoLoop, we believe that there will always be a role for the human researcher, albeit likely a different one from that of today – and tomorrow. Tech entrepreneur Santiago Valdarrama says, “AI will not replace you. A person using AI will.” We wouldn’t go that far, but we certainly agree that qualitative researchers and insights professionals can view AI as a tool that can enhance both efficiency and effectiveness.

Although the conversation about AI in qualitative research has been dialled up to maximum volume in recent months, it’s not all that new.  Industry experts, such as Mike Stevens, have been writing about the use of AI in qualitative research for many years. Early tools used basic algorithms to analyse and categorise qualitative data, such as interviews or open-ended survey responses. As AI technology advanced, researchers began to use more sophisticated techniques, such as natural language processing (NLP) and machine learning, to extract meaning from text data. These techniques enabled researchers to identify patterns and themes within qualitative data sets, improving the speed and accuracy of qualitative analysis. It’s the recent advances in generative AI – and the speed of development – which has turbo-charged the conversation, and with it, the concerns that many researchers are experiencing.

How AI is changing qualitative research 

Today’s new tools are changing how we do both data collection and analysis:

AI and qualitative data collection

AI enables the extraction of insights from sources such as focus groups and depth interviews, as well as from ‘qualitative adjacent’ sources such as social media, online forums, open-ended questions in online surveys and other web-based sources. Some tools are designed to capture the data directly whereas others take inputs of data collected elsewhere, such as transcripts of focus groups. For example, CoLoop is designed to seamlessly upload transcripts or video content together with names of participants, interviewers or moderators, to facilitate the analysis process.

However, it is in data analysis that generative AI has the biggest potential to disrupt the industry by helping researchers identify key themes and patterns, reducing the need for manual coding and analysis, and making the process more efficient. Human researchers’ skills and experience will still  be essential when transforming this information into valuable insights for finished client decks and reports. To this end, AI will act as a co-pilot for researchers to assist in their analysis, not as a replacement to fully automate their jobs.

AI and qualitative data analysis

Technologies such as NLP can make sense of language in much the same way that a human would – but much faster. Sentiment analysis is a type of NLP, which is used to understand the emotional tone of a text. For example, as well as uncovering the key themes that are discussed in a focus group, the language used can be revealing of underlying emotions towards the products, services, issues and topics under discussion. Topic modelling is another technique, used to identify patterns in large datasets and can also help pick up themes and issues. This sort of analysis is particularly useful when comparing data across multiple focus groups or online communities. Overall, AI takes the hard work out of the labour-intensive task of reading, categorising and making meaning from vast quantities of conversational data. But human researchers need to make sense of the work that the AI had done.

Implications for researchers of using AI in qualitative research

The use of AI offers a wealth of opportunities for qualitative researchers and insights professionals. There are challenges too to be overcome, but we believe the rewards are worth the work.

Opportunities for qualitative researchers and insights teams

Productivity: one of the biggest challenges of analysing qualitative research is the volume of data. This is especially the case in bigger projects where there are multiple interviewers and moderators or multiple markets or audience sections. AI tools can automate and accelerate data analysis, which frees up researchers' time to focus on the interpretation and application of results. 

Cost savings: on the agency side, automated analysis and increased efficiency mean putting fewer people onto projects, making them more efficient – this can help agencies become more competitive without compromising margins, and make research affordable for smaller clients. On the client side, this can open opportunities for insights professionals to do more analysis in-house. 

Better insights:  AI can help researchers identify patterns and themes within qualitative data that may have been overlooked in traditional qualitative analysis methods. Client-side insights professionals can use AI to take reports delivered by agency partners and add additional insights, or to analyse multiple projects from different suppliers. AI tools can also help by bringing client-side stakeholder teams into the process, which can help with the communication and activation of insights.

Fewer assumptions: Researchers understand cognitive biases better than most people. But we are still susceptible to them. Whilst AI tools may have biases (see below), they won’t be the same as researchers’ own preconceptions and unconscious assumptions. By uncovering insights that challenge these personal assumptions and biases, researchers can arrive at more objective and accurate conclusions. 

Challenges for qualitative researchers and insights teams

Potential biases: AI tools are vulnerable to potential biases arising from the data they are trained and supplied with. Biases can lead to results that may reflect stereotypes present in their training data, making it necessary to sense-check the output of AI analysis. Although providers of AI models make extensive effort to re-align models from biases, researchers must ensure that the results of AI analysis are never considered as the final output, but instead use them as building blocks to construct a comprehensive narrative.

Accuracy and reliability: generative AI is able to synthesise content based on the data supplied to it of varying originality. It can be used to create completely new content or extract existing snippets from supplied text. For example, if you use generative AI for desk research, you still need to fact-check and provide sources to back up your assertions. If you are analysing qualitative research content, the output is based on your transcripts, video or audio, so there is less room for error but, as above, human input is essential.

Training and learning: it is tempting to think that these new tools will deliver perfect results straight out of the box, but, like any tool, there is a learning curve. Researchers must be willing to invest some time to learn what is possible, and to conduct iterative experiments to understand when AI can be helpful, how and when to combine human and AI input, and how to maximise the output from AI tools.  

Changing roles: with the automation of some tasks, such as data collection and analysis, the role of the researcher may shift towards the interpretation and application of results. This requires researchers to develop new skills, such as data visualisation and storytelling, to effectively communicate insights to clients and stakeholders. As AI enables more qualitative research to be conducted in-house, client-side researchers must work collaboratively with other departments, such as IT and data science teams to ensure that the insights generated through AI tools are integrated into broader organisational decision-making processes.

What’s next? Future developments in AI and qualitative research

The next few years will see increased automation of qualitative data collection and analysis, further integration of AI with other research methods and development of new AI tools and applications. At CoLoop we are proud to offer one of the most advanced tools for qualitative research data capture and analysis currently available – and we aim to stay ahead of the market by building in new developments as the technology improves.

Specifically, advances in NLP, sentiment analysis and machine learning will continue, enabling tools to better integrate the context of language and interpret it more accurately.  Additionally, automated transcription and translation will improve, further driving down costs and timescales for qualitative projects. We can expect to see virtual research platforms that use AI to facilitate group discussions becoming more widespread, and we will even see AI-generated research participants. And that opens up a whole new set of challenges and opportunities.

Conclusion

So, although AI won’t be replacing human researchers right now, the technology is not going away. Co-founder of Netflix, Marc Randolf has said that “if you are unwilling to disrupt yourself, there will always be someone willing to disrupt your business for you.”  There is huge value for researchers in integrating AI into current practices, alongside all the skill, experience and expertise that only humans can bring. We understand how unsettling it can be to go through technological disruption, but we truly believe that for qualitative researchers, AI can be a support rather than a source of concern, a turbo-charging superpower and a co-pilot on the journey through our changing world.