Bots have been hailed as the key to transforming the customer experience, but Jasmeet Sethi, head of operations at Ericsson Consumer Lab, warns that there’s a downside for market research.
“We’re moving away from an app economy to a bot economy,” Sethi told the MRMW Europe conference in Berlin. “We already see signs of AI taking centre stage, we already see signs of AI displacing the smartphone as the possible primary user interface.”
Computing, he noted, is moving “from the palm to the face or the body” as voice-based devices enter the mainstream - and that has implications for the market research industry.
There are gains to be had from using bots for everything from automating dispatch of email reminders to administering polls and surveys. They could even find a role in qualitative research, Sethi suggested, prompting discussions or diary studies. “There are a host of different applications where we can see how AI-driven bots could transform market research,” he said.
But, equally, there are challenges associated with this development. “What if, as an end user, I decide I’m going to build a bot that can possibly take surveys for me and earn some money?” Sethi asked. There are a lot of people thinking that way, he stated, and the process of building such a bot can be remarkably easy using drag-and-drop features on some websites. “I can go in [to a website] and just with about four steps create a bot that actually mimics my personality and engage on platforms like Facebook and WhatsApp,” he reported.
And, he added, such bots are already starting to take surveys. At a Stockholm hackathon, his team had created a bot that studied the patterns of how humans take surveys - time taken etc - and which was then able to “automate a bot that can take a survey exactly like end users”.
It’s a scary development, he suggested, and one that is giving him sleepless nights. While it’s not a burning issue right now, that may be very different in the near future, especially given the speed at which AI is advancing. “I can only imagine how many developers out there will come up with these sort of bots that are really going to disrupt our industry and contaminate data.”
Who knows what is the level of data contamination in global surveys, he mused. And whatever it might be now, it is going to be “very significant” soon. “I don’t really see the industry working on this aspect,” he said. “We have talked about data quality a lot but we haven’t talked about data contamination in the context of AI.”
To put his concerns into context, he observed that bot traffic has overtaken human traffic on the web: 52% of all web traffic is bot traffic. “If you are an MR supplier doing these unmoderated recruitments online on different platforms, no wonder some of these bots are actually making it into your panels or are starting to take our surveys.”
These bots are intelligent, he added, being quite capable of bypassing security measures like Captcha and earning money without any user intervention. “This is top of mind for us,” he said, “because we would really like to see some players in the industry launching a data-driven counter-attack.”
Sethi held up as an example the work of two students at UC Berkeley who have apparently managed to do what Twitter cannot, or will not, and come up with a way to identify and expose propaganda bot accounts on the platform.
“They’re studying simple patterns which are all about when are these bots tweeting - at what point in time and with what frequency - what subjects, how many followers on average and comparing it with the other Twitter population out there.”
If a couple of students can do that, then how much more should the industry itself be able to achieve with all its talent, he demanded to know. “I think this is the next business idea,” he added, “because in the next five years you would see a rise perhaps in our industry of firms that would be doing what these students are doing today - which is sifting away bot traffic from human traffic for us as clients. And that’s the real opportunity I think we should be leveraging.”
