By Johan Jeuring

One-to-one communication skills are essential in most professions. A doctor, veterinarian, and most other professionals need to appropriately communicate with clients. Many university and vocational programs require communication skills, and train their students therein. Communication skills are also often trained on the job, at any level: from management (how to deliver bad news), to call centre operators (how to convince people to buy a particular product).

Communication skills are best learned through practice, in role-play with an actor or with a simulated patient. Training one-to-one communication skills requires at least three people: two people engaged in a conversation, and a third assessing the conversation and skills. Such trainings require a substantial amount of time and organisation. The last decade has seen the development of a number of digital environments that support training communication skills. A student performs a conversation with a virtual character, or with a character that appears in videos, and the learning environment assesses the performance of the student. Usually these environments do not replace the traditional real-life trainings, but they are used to better prepare students for such training sessions, or to offer training possibilities in situations where it is hard to arrange real-life training sessions.

A number of games developed in the RAGE project train learners in communication skills. I have been involved in the development a communication skills game at Utrecht University for the past three years: Communicate! The question we discussed most during the development of Communicate! was: how can we best develop real-life scenarios for such a game? Developing a scenario involves translating a communication model to a realistic scenario, including common errors learners make, together with the realistic responses of people on such errors. Each choice made by a learner leads to a (possibly negative) score on one or more communication skills, and to an emotion and a response from a virtual character.


We developed Communicate! in a team consisting amongst others of six communication skills teachers or trainers. A decision we made very early on in the project was that we wanted the communication skills teachers to develop the content for the various scenarios. Looking at the number of scenarios produced for Communicate!, more than 200 by now, that decision has been a fruitful one. It also ruled out the use of artificial intelligence: the communication skills teachers were not well-versed in programming techniques to produce the necessary formal structures for an artificial intelligence approach. Furthermore, the teachers did not trust anyone else than themselves to produce correctly formulated sentences.

Reflecting on our experience from the past years, I want to discuss three aspects that I found interesting or surprising, related to scenario development.

The first aspect is about what you can learn using a game like Communicate! The most obvious kind of global learning goal is to learn protocol-like communication skills, such as delivering bad news, handing out medicine to a patient for the first time, or performing an anamnesis, and to experience the kind of reactions you get from people when you make the wrong kind of remarks. In these kind of scenarios there are good and bad choices, and we can distinguish scores on particular learning goals. Another use of Communicate! is to learn more complex communication skills. For these skills, the scenario does not offer clearly good or bad choices, but it offers choices that spark a discussion amongst students. This use of Communicate! requires multiple students to play the game together. Finally, a third use of Communicate! is to learn how to translate communication models to real-life scenarios. Here a student does not play a scenario, but instead develops a scenario. Only after we had developed many scenarios ourselves, we realised that this activity would be useful for students too. Once a student grasps basic communication skills, developing real-life scenarios requires higher-level communication skills.

The second aspect is related to offering the possibility to follow various paths through a scenario. Communication models often prescribe a particular order in a scenario, but in real-life it is often hard to adhere to this order. For example, in an anamnesis, a doctor should ask questions about several aspects of an illness: duration, location, and more, and the anamnesis model prescribes a particular order. But in many real-life discussions, a doctor and a patient do not discuss these aspects sequentially, rather they interleave questions and answers in various orders. Explicitly modelling all these orders in a scenario leads to an explosion in the so-called `dialogue-trees’. Our solution to this problem is to allow the construction of dialogue trees on particular themes, such as for example the duration of an illness and the location of an illness, and then combine these separate dialogue trees into a single scenario in which parts of these dialogue trees may be interleaved in various orders.

The third aspect is the final score of player after playing a scenario. Ensuring that the scoring in a scenario makes sense is pretty hard for more complex scenarios. For example, if a student takes many wrong steps, scores on some of the learning goals might add up to the maximum possible score for that learning goal, although teachers assess the performance negatively. To check if the scores make sense, we might try to calculate all possible scores for a scenario, and show these to the scenario author. With the presence of the interleaving combinator introduced in the previous paragraph, the number of possible paths is potentially exponential in the number of statements, and showing all of these to an author is infeasible, both from a computational as a usability perspective.

We will think more about this, and any input is welcome.

Posts are written and signed by its authors. RAGE Project does not responsabilize for the opinions and comments published in them.