How many times in the past month have “ChatGPT” and “AI” come up in conversations at work or with friends? If the answer is zero, please invite me to the rock you’ve been living under. It sounds like a great place for some peace and quiet.
It’s difficult to avoid conversations about the promise and peril of Artificial Intelligence. No one seems to call it by it’s full name anymore because the acronym is so ubiquitous. And to be fair, AI is nothing new; a lot of smart people have been building the AI ecosystem for years.
However, it’s the explosion of ChatGPT in recent months that has brought the conversation to the public square. Even the consumers we’re so eager to learn from are talking about it. The conversation is now at the water cooler, on Zoom calls, and in the boardroom.
I’ve been in some of these conversations and have closely attended to how those in the Marketing Research industry are reacting. Much like broader responses, they run the gamut from fear and paranoia that AI may undermines the foundation of the industry to irrational optimism that it’s a panacea for the industry’s many plagues.
The robots are here…and they speak like us!
Unsurprisingly, early conversations have touched on the impact of AI on market research fraud. Data quality is already top of mind in the industry, and many have been burned by fraudulent participants and bad data. But what’s that? Here come the bots… and they’re smarter AI-bots? Everyone run for shelter before it’s too late!
There are very legitimate concerns around how survey participants could use ChatGPT and other tools to generate very compelling, yet fraudulent, open-ended responses in both short and long form. And with open-ended responses continually and heavily leveraged to identify what data is “bad” and should be discarded, this should be a topic of conversation. I’m not dismissing the magnitude of the challenge, but we can fight AI-empowered fraud using the very same technology.
When I first saw ChatGPT blow up, I flashed back to my days as a professor and was thankful that back then… I didn’t have to face the near impossible challenge of detecting an onslaught of papers written by robots posing as students. But then in a matter of weeks, an entrepreneur developed a way to detect if ChatGPT was used to create content.
Is that battle over? No. But it’s an example of how quickly we can leverage AI for good, rather than evil. That same thing is already happening in market research. Companies that don’t currently have AI tools at their disposal to detect and address open-ended data quality problems will very soon have them. With the barrier to entry dramatically lowered, we should relish it as an opportunity.
If the robots speak like us, can they really chat like us?
It hasn’t taken long for people in market research to also ponder how AI might usurp the role of a researcher. How can we use AI to do work typically done by a trained research professional?
Can a bot conduct qualitative interviews? On some level, yes. And again to be fair, several companies have been doing this for years – and getting continually better at it. Generative AI accelerates and democratizes access to tech, enabling a conversational interaction between humans (research participants) and robots (moderators). While online discussion boards and online qual are likely better moderated by a human, we are fully into the human-assisted moderation world already.
Can a bot write questionnaires? Yes, the results aren’t great, but I’ve certainly seen worse. Because of the relatively predictable structure of questionnaires, AI tools are able to respond well to prompts, guiding creation of a questionnaire consisting of a certain number of questions and focused on a certain topic. Most of the questions are well constructed. The answer choice sets tend to be a good start, if also imperfect. However, question ordering to mitigate bias really doesn’t yet seem to be AI’s forte. This said, AI tools can very quickly craft a viable survey instrument that is mostly better than a survey someone with little to no research experience might construct.
While we can be rightfully critical of how out-of-the-box AI-based tools work for crafting research documents, we need to remember that the models aren’t yet trained for that. They will be. Opportunities for AI-assisted survey design and execution, qual and quant, is happening now — and quickly.
The robots are coming for my job!
Are you sitting back wondering if AI will replace you? If yes, and that’s all you’re doing, maybe you should be worried. But for those skilled in design, moderation, analysis, strategy, storytelling, and more , you can also look at AI as an opportunity rather than a threat.
Remember just a few years back, when conversations around automation and DIY furrowed brows about jobs being automated away? Did that happen? No, not really.
I remember not too many years ago when I spent hours upon hours coding open-ended responses manually. Now, there is no way I would do that without the assistance of automated (and perhaps AI-driven) tools. Let the robots take the work they’re good at. And let me keep the work I’m good at as a human.
Researchers have by and large embraced tools and platforms that speed up their work and automate tedious and manual processes. The sophistication of AI further accelerates this move to spend time where time is well spent, instead of rote activities.
So what does this all mean for researchers and their employers? It means you need to quickly get up to speed. If you’re not exploring generative AI, –- actually using it and not just reading about it, you’re late. But the good news is that the barrier to exploration is almost non-existent. There are free and low-cost AI-based tools aplenty. This space is moving incredibly quickly, but you can also catch up quickly.
Here are four incredibly easy things you can do this week to dip your toes in AI. First, start by signing up for the free version of ChatGPT through OpenAI (or choose another platform). Then…
- See how well it writes a survey by using the prompt, “Design a 15 question survey about [insert topic].”
- Code open ends by pasting in a set of open-ended responses and using the prompt, “Categorize this list of open-ended data.”
- Level up coding open-ends by using the same set of responses and the prompt, “Create a dataset that categorizes the following as positive, neutral, or negative sentiment in .csv format: Verbatim | Sentiment | Reason.”
- Pull out themes from qual data by pasting excerpts (or full text) from an interview and using the prompt, “Identify and define themes from the following text, organize them into a bulleted list, and use direct quotes from the text to provide examples of each theme.”
Playing with these simple prompts should take about a half hour, and while basic, will give you a sense of what the technology will unlock for those in the research space.
Resisting generative AI is truly futile. Explore how AI can improve your business by staying open minded, identifying opportunities, and getting real hands-on keyboards experience.
This article was originally posted on GreenBook here.