The first chatbot I ever came into contact with was Cleverbotwhen I was in school. Cleverbot is a bot created by AI scientist Rollo Carpenter and first went online in 1997; anyone could have conversations with the bot through the little text box in the centre of the screen, and each time the bot would store the human responses in it’s databanks to learn from.
By the time my friends and I had discovered it, it had already gone through millions of conversations with internet users around the world, and thus had gathered a massive dataset of conversation pieces. But it still didn’t carry on a conversation like it should; if I were to use an analogy, it was somewhat like talking to someone with short-term memory loss, the bot would not remember what conversations we'd already had, and drop the topic of conversation in favour of another in an instant, and wouldn’t retain any information long enough for a truly personalised encounter.
However, Carpenter has since created other bots based on the Cleverbot algorithm, and this software has developed to the point where its ‘Evie’ persona has passed the Turing Test, fooling 59% of its users into thinking that they were talking to an actual human.
Now, the chatbots we encounter on a daily basis are obviously not as advanced as this, they’re programmed with set responses for more specific conversations; but it does make one wonder about the current and future applications of chatbot software in the insight industry.
Even though chatbots are becoming more of a common feature on websites and in social media communication applications, 60% of consumers preferring to wait in a queue to talk to a human agent over interacting with a chatbot, which brings us to the all important question, are Survey Chatbots coming of age, or are they ageing out of all usefulness?
Coming of Age - The State and Use of Chatbots
The premise of using chatbots for surveys is indeed an intriguing one; a device that participants can interact with in real-time, gauging their intentions and sentiments of a website, brand, or service as they’re using it can be invaluable. But have we developed them to the point where we can take full advantage of the opportunity they present?
Currently, the interactive element presents an engaging opportunity for both participants and the insight professionals who use the data. While surveys are somewhat interactive, the user interface of a chatbot allows for more of a give-and-take, conversational style survey that allows participants to relax a little more in the informal environment. And depending on the algorithm you use, the chatbot itself can formulate and customise the next question depending on the answer given beforehand. Response rates will improve with this type of non-traditional style survey, as it allows a relationship to form between the respondent and the chatbot, and by extension, the researchers/brand conducting the survey.
Chatbots now aren’t just website features, they can even be incorporated into messaging apps such as Facebook Messenger, WhatsApp, Snapchat, and more to reach a wider audience that doesn’t generally read emails outside of work or answer their phone unless they know who’s calling. This wider integration is of a massive benefit to insight professionals, as we now have better access to generations on their own terms.
Typically, survey chatbots are used for collecting data for user metric analysis and lead generation, but there are survey chatbots that have been developed (or are currently in development) for more specific research purposes, that can be embedded into other communication-based software tools for ease of distribution and participation.
But while chatbot capabilities have grown, another important question to consider is how far do consumers trust chatbots? Because if the trust isn’t there, then consumers won’t use them at all. According to a study conducted by Userlike, we are willing to chat with a bot for simple enquiries, but we like to know for certain that it is a bot we’re interacting with; when they start acting like humans, it’s seen as deceptive and the trust in them reduces significantly.
This sentiment, while seemingly picky, is actually rooted in a need for transparency and equal footing. With experiments into bettering chatbot algorithms and the language they use, it seems realistic that it could get to a point where we’re not sure whether we’re talking to a human or an algorithm, this layer of unpredictability is something consumers find deceitful on the company’s part.
What is the Future of Survey Chatbots?
This more interactive and immediate way of delivering surveys in realtime has the potential for generating some brilliant insights if the surveys are simultaneously targeted and generalised enough. But this potential is fully linked to opening the chatbots up in a safe way to consumer data, analysing human answers and learning how to personalise and humanise themselves more to anticipate reactions, respond to queries and responses as we would as humans, and thus become more helpful in the survey data collection process.
We also need to take into account, whether they’re going to be used as common research tools, so the time and effort we invest into them will be worth it. At the moment, it’s primarily the younger, more tech-savvy generations who interact with chatbots, while the elder generations still use email and telephone. Depending on who the target audience is, chatbot research tools might not be useful at all no matter how developed they become.
With conflicting opinions on the merits and future of chatbots, some believing they’re annoying little pop-ups who only interfere and distract from the purpose of the visit, and others seeing the potential they hold as real-time data gatherers for insight professionals to factor into their analyses, we find ourselves with a few paths to take:
Is where we take the plunge and allow survey chatbots to create more personalised encounters with participants, basing their next questions on their answers. This is going to be a tough ride, as we need to make sure that this data isn’t tampered with, isn’t biased, and isn’t corrupted like Microsoft’s Twitter bot Tay. The most extreme case of chatbot corruption, Microsoft’s Tay, was a Twitter bot that learnt as it chatted with other users, was utterly corrupted by them in less than 24 hours.
Is where we find other ways of trying to personalise chatbot interactions without opening up the algorithm to participant interference. This might be from extensive development in-house, with more datasets to broaden the range of answers the bot find’s appropriate given the parameters of the survey. While we have a semblance of control over survey chatbots at the moment, if we want to allow for more customisation and personalisation in its answers, then we need to be sure that it won’t be corrupted by researcher biases as it learns.
Is where we find that chatbots have run their course, they’re as useful as they’re ever going to be, and while they’re great tools to use while we have the platforms to support them, there will be other interactive survey tools that come along and replace them soon enough.
I do believe that survey chatbots have yet to reach full maturation, and there is some way to go yet before we can use them to their full potential. But while there is a need for more interactive research and automation processes to take some of the pressure off researchers, survey chatbots will always be a viable tool to use, so time is something that we have to dedicate to developing chatbots into the most useful iterations of themselves for use in market research.
As a graduate of Creative Writing, Emily has a passion for content creation. She brings our global vision to life through her excellent writing and editorial skills across a broad selection of our content, and manages communication through social media channels. You can follow her on Twitter and connect with her on LinkedIn.