Improving communication skills has benefits in almost any job, but they're expensive skills to practice—typically requiring live role play or expensive technology. But there is a cheap, easy-to-use tool available to create conversation simulations for training.

Twine is a simple, free, open-source tool for creating non-linear, interactive text stories in HTML. It's intended for games but it has huge potential for scenario-based training and simple chatbots. In this interview, Paul Bills talked about how learning professionals can use Twine to create interactive training on almost any conversation scenario.

BB: How do you use Twine when you're building chatbots for conversation simulation?

PB: I use Twine as a kind of first level. Twine doesn't make chatbots according to a lot of people's definition of chatbots but Twine can make an experience very similar to any of the full-blown training chatbots I've worked with where the user chooses options in a scenario.

That's a particular kind of chatbot, and Twine is really built for that approach. Since I had made some chatbots along those lines, I remembered Twine and went back to it and came up with this kind of chatbot where the chatbot is presenting a scenario and asking a user to select an option in that scenario to see how the story plays out.

That's exactly what Twine is built for. While some people would probably take issue with me saying to use Twine to make chatbots, that's really what I mean by it. Twine can make this kind of scenario-based chatbot where the computer is "talking" to you and you are selecting what to say back. So it's perfect for that specific use case.

BB: Would you recommend scripting or storyboarding the interactive stories for scenario-based communication skills training with a chatbot?

PB: Yeah, definitely. I don't have a specific recommendation of how to do it or which method—scripting or storyboarding or any other kind of planning method—but definitely you want to plan it out and make sure it's creating the experience you want to create. Where I think it's most effective in communications training is letting the user be surprised by outcomes. Simulations become fun when they let you fail, even when you do that on purpose to see what would happen.

Have a plan for all scenarios, including those where the user purposely fails to see what happens but also where they're surprised—when they think they're picking something right but maybe they know it might not be the totally right answer.

Plan for these scenarios whether you do that in storyboarding or through scripting. You want to go in with a picture of, not just the core conversation but those branches as well, and where it might lead, where it might go.

BB: How complex does your branching get in these simulations?

PB: That's really up to you. Twine as a tool isn't limited by the number of branches. It's up to how resource-intensive you can be; how much time and how much planning you want to put into it.

That being said, there are natural limits to what most designers are able to do given the amount of time they have. What we found in building chatbots is that you generally want at least three core branches: the fail state; the just getting by; and the great work. You want at least those three scenarios to keep it interesting.

Those are the kinds of core branches and within those you could start heading down the negative path early and make choices within that path. And then maybe if you make right choices within the negative path, you get back to the neutral path, and then to get back to the good path.

But again, that's all just dependent on how much time you have to plan out and build the complexity of your own story, but as a tool Twine is really limitless that way. It's as much as you need; it can be as complex as you want.

BB: Are the bot responses in text, or do you do use professional voice talent or the developer's own voice?

PB: Actually most of the ones I've done are purely text. It's just reading the text back and forth. You can add voice and in that case, if you're competent enough, I would say go for it. Doing the voice yourself keeps that budget low. But if you're going to add voice, you don't want the voice acting to take away from the experience. So, voice acting that's not as confident or as clear might actually pull people out of the story you're trying to create, and then they start focusing on the quality of the acting itself. So in most cases I would recommend no voice at all—just the text interaction because then people can fill in the voice in their own head and that helps them stay immersed in the story.

BB: How much communication skills training can chatbots carry?

PB: Yeah, that's a good question. It's one we continue to explore in my work. Generally, especially at first when you're starting out, this is a supplement to some other kind of training. So maybe you watch a video, and then afterward the training application says, "Hey, let's practice what you learned in that video, in this chatbot conversation." And then you can try several different scenarios relating to what was just taught in the video, or in an instructor-led class.

Some of the more interesting use cases for it would be pre-work, where before someone comes to a formal training session or before they take an eLearning of some kind, you ask them to try out some scenarios. A popular use case for this is for new managers—people who were great individual contributors before, now being promoted to a manager. Suddenly they have to deal with all these scenarios that they weren't dealing with before. This can be a great way to introduce them in a low stakes way to those kind of scenarios. You give them a conversation simulator made in Twine that asks them to deal with someone who's always late, or deal with someone whose quality of work is slipping. You can do that as pre-work before a formal training to help them appreciate what the stakes are here. If they go into that simulator and they see that someone decides to quit their job because you didn't handle it right, or you see someone's struggling because you're not giving them the proper coaching. They can see those effects. In that low stakes way, where you know no real people are being affected, that can be a great intro before a formal training about communication skills.

You could have simulators like this, with some extra explanation right within the program, and deliver some training with text and images. Then the user has to make a choice based on what they just learned. You could have that kind of back and forth, and do all of the training within the scenario itself, and then I would probably recommend at least some kind of group discussion afterward to help digest what was learned and what happened so that they understand how to translate what they just did there into the workplace, and it's kind of cemented in. Rarely is it the entire experience but it can be a really great supplement.

BB: So it sounds like the context of this is generally in a group setting, whether say in an online situation where everybody is at remote locations and connected, for example, using Zoom, or in a classroom setting.

PB: It's usually where there's a cohort of people all taking some kind of training together. And then, when they break away, then you introduce something like this to keep them practicing, keep them thinking about that.

BB: That answered one of my questions, which was how do you handle the feedback? But if you're in a group setting or in a cohort, then that's taken care of.

PB: I want to encourage people to experiment with this. I've seen similar kinds of trainings out there. I think anyone who's willing to experiment with it will learn a lot very quickly. That's really my whole goal.

Want more?

Paul Bills will expand on the use of chatbots in his session, "Build Your Own Chatbots for Conversation Simulation with Twine", April 23, in The eLearning Guild's L&D on a Shoestring Online Conference.

In this session, Paul will show you:

  • How to use Twine, a free, open-source interactive storytelling tool
  • Best practices for conversation simulation design
  • Other chatbot tools that can expand your conversation simulation and training
  • How chatbots can support training
  • How to export Twine stories to HTML for easy integration into your existing learning ecosystem