3 Ways in Which Artificial Intelligence Can Complement Qualitative Research
Back in 1950 Alan Turing’s famous question “Can machines think?” opened his paper on ‘Computing Machinery and Intelligence’. Jumping ahead in time somewhat, the advancements around artificial intelligence (AI) are manifold and span pretty much all aspects of our lives. AI is serving millions daily – be it via your smartphone, your car, your bank or your house. Sometimes you actively use a service such as Siri or Cortana, other times it’s less obvious and very often it even goes unnoticed. The AI wave is having an impact on industries across the board including market research.
“Artificial Intelligence is not a Man versus Machine saga; it’s in fact, Man with Machine synergy.”
AI is inherently based on mathematical models and driven by numerical data. You might assume that this clashes somewhat with the descriptive nature of qualitative research. After all, qualitative research is concerned with a set of methodologies designed to explore; to delve into the depths of human behaviour and its reasoning. In order to achieve this qualitative research questions tend to be contextual, broad and non-static. Moderation is often described as intuitive and the data output – text, audio, visual and audio-visual - ‘messy’ and immeasurable. Today however, AI products have advanced beyond mere numbers. They are now equipped to deal with this ‘messy’ immeasurable content, or at least assist in the process.
MROCs are a great example of a market research approach which provides vast amounts of qualitative data output. The sheer volume of community member content produced over weeks, months or years is a challenge for any community manager. Text and sentiment analysis are already in full swing in qualitative data analysis, but there is another exciting application of these tools in that of insight community engagement.
By applying deep learning to a combination of text and sentiment analysis, login, activity and profile data, market research can leverage AI to anticipate member disengagement before it happens. This enables the (human) researcher to manage member engagement proactively, to reduce churn without firefighting. And there is nothing to stop this clever AI technique from being extended into engagement optimisation, for deep learning to ‘learn’ what engages who, when and where. All leading to higher quality feedback for both researchers and clients.
2. Conversational Insight Activation
We have already seen variousmarket research dashboard applications enter the arena, all designed to provide internal stakeholder access to real-time data in a visualising appealing manner, i.e. to encourage insight based action. But what if we took it one step further… beyond the dashboard to an insight activation chatbot.
For an entry level insight activation chatbot, i.e. one that deals with quant data only, the AI required is still twofold:
The analytical ability to identify correlation within a vast number of differing quant data sets
The natural language ability to articulate this correlation, with supporting visuals
Working off the back of a dashboard style database, an entry level bot of this nature would be able analyse both historic and current data, and summarise only query relevant insights in bite-size easily digestible qualitative chunks. Individual customer persona chatbots could even be programmed to emulate Q&A style conversations with business members. Nowweare using qual to embed insight within organisations, that’s pretty clever.
If we were to add deep learning into the mix we could take our chatbot even further. This would allow for more creative proactive use. Your friendly insight activation chatbot no longer has to wait to be called upon; rather it can push insights out to company members. Recipients feedback on relevance when they are able and the bot learns what is beneficial to their job role, at what time, and even the language to use to resonate. It’s got insight culture written all over it… With one small caveat. Each bit-sized insight would have to come with full study links. Even when qual research data is added to the quant database, a chatbot identifying causation? There are a few too many holes intext and sentiment analysis for that from my perspective, at least as it stands today.
Advances in both natural language and voice processing have seen the use of virtual assistants, both commercial and personal, taken to a whole new level. Commercially this form of AI can be applied to customer, information and entertainment services as well as purchasing functions and has become increasingly popular with the big players such as Amazon, Facebook and Google.
In our blog,5 Big Qual Market Research Chatbots You Can Build Today we proposed a specific use of a chatbot for moderation assistance in big qual. But there is also the potential to create a generic research assistant. As in the commercial scenarios referenced above, the ‘Researcher’s Assistant’ would provide participant information services either full-time, or out of hours. It would effectively ‘staff’ the research project. In the current AI circumstances this staffing would be limited to natural language processing and generation based on a knowledge bank but nevertheless, the RA would provide a degree of interactive support to bolster participation where it was previously a challenge to do so, i.e. evenings, nights, research projects spanning different times zones, and / or free the human researcher to focus on the more complex cognitive tasks.
Market research organisations and professionals are embracing AI with open arms, trying out new approaches and research set-ups. It is envisioned that researchers and AI will be working alongside each other to create (qualitative) insights that previously were beyond the realms of possibility. AI is here to stay, let’s get involved proactively, let’s lead the way.