How do buttons influence UX?

Buttons help us steer the conversation and limit the ways our users can respond to our Virtual Agent (VA). This, in turn, reduces the chance of users going off-script, and guarantees an answer from the VA. But using buttons everywhere may not be desirable, either: you want your users to be able to use their own words, too. So when you design a conversation flow, you have a choice to make: use buttons, or not?

Rather than making this decision on gut-feeling alone, let's take a look at what research has to say!

In this post, I'll share insights from two research papers: "Understanding the user experience of customer service chatbots: An experimental study of chatbot interaction design" by Haugeland and colleagues, and "Chatbots as Advisers: the Effects of Response Variability and Reply Suggestion Buttons" by Milana and colleagues.

Let's dig in.

How do buttons impact conversational experience?

From our point of view as conversation/flow designers, buttons are a neat design tool. But how do our users experience these buttons?

With that question in mind, Haugeland and colleagues set up an experimental chatbot conversation using the boost chat panel. In one group, participants could only respond by writing ("free text"), while in the other group they could both write text and use buttons to reply.

They found that the participants that could use buttons to reply had a better conversational experience than the ones who could not. They reported better scores in terms of how pragmatic/practical the conversation was, as well as how enjoyable the conversation was to them (something the paper labeled "hedonic experience").

For example, when asked to rate their conversations from complicated to easy, participants that could use buttons rated them as easier than the participants who couldn't use buttons at all. That's a clear benefit of having buttons!

How do buttons impact VA perception?

Buttons can also influence your users' impression of your VA. For example, when you've explained your problem to a human service operator, their response will influence what you think of them: how capable they are, how helpful or how friendly, to just name a few things. When they provide you with relevant suggestions, this signals competence and understanding - and generally improves your impression of them.

The same may apply to your VA, where buttons can be those relevant suggestions. To test this, Federico and colleagues investigated what happens when a VA gives advice with, or without using buttons.

In their setup, participants were asked to make financial trading decisions. To inform their choices, they had access to a newsfeed with information as well as a virtual trading assistant which would give them advice on what to do.

The VA and the newsfeed would sometimes provide conflicting advice. For example, the VA could say "I believe Paulinho's portfolio will increase", while the newsfeed clearly shows a prediction for this portfolio to decrease in value. In those cases, Federico and his colleagues noted how often the participants would follow the VA's advice - and how the use of buttons influenced that decision.

They found that, when buttons were used, participants were more likely to follow the advice from the VA over that from the newsfeed than when no buttons are used. They suggest that this is because the buttons make the VA look more competent - and advice from a competent source is more likely to be followed.

Downsides

So far, we've only seen the positive effects of buttons in a conversation. They increase the conversation experience, and likely make the VA look more competent - but not always.

For example, Haugeland and colleagues point out that the buttons need to be relevant to the conversation at hand for them to be a positive contributor to the conversation. In other words: simply blasting buttons everywhere is not going to make your conversation better. More specifically, the buttons need to help the user forward:

"...for example, to help the user refine a request or to indicate relevant options" (Haugeland et al. (2022))

Next to that, Federico and colleagues point out that using buttons to boost a VA's perceived competence can also backfire. Buttons may cause your users to rely on your VA's suggestions too much - or, in other words, stop thinking for themselves. Best case scenario, this means they may stop writing messages and abandon the conversation if the buttons don't show them a clear path forward. Worst case, users may base important (life) decisions on your VA's buttons - which can be dangerous in some situations, such as healthcare settings.

To button or not, that is the question

In summary, buttons remain a great conversational tool. For starters, they help us steer our users and provide suggestions to help them wherever they may get stuck. It is not always easy to formulate a question, and buttons can be a great support to do exactly that. Not just that, buttons can also help make your VA look competent, improving the chances of your users accepting its answers.

But don't overdo it: buttons that are not relevant will likely not get you the improvement you're after. Worse, they may make your VA look dumb and make your users feel as if they're stuck.

So the next time you're considering using a button, ask yourself: does this button show insight into the conversation? Does it help the user forward, or show alternative options? If the answer is yes - then, by all means, button away!