Draft Choices and Alternate Paths with ChatGPT

In a previous blog post, I explained how to start writing branching scenarios with AI tools like ChatGPT and Bard. In this post, I’ll continue with that same scenario, focusing on how to draft choices and alternate paths with ChatGPT. Coming up with alternate paths and plausible wrong answers is always one of the most challenging tasks in writing branching scenarios. Even with good analysis and questions to focus your planning on the mistakes and consequences, it’s fairly common to have some gaps where there isn’t a clear wrong answer for a decision point. In some cases, I think ChatGPT and similar generative text tools will be able to help fill in those gaps in scenarios.

As with my previous post, I want to set expectations. This is not a post from a self-proclaimed AI expert with a list of magical prompts that will fix all your problems and instantly give you 15 hours back every week or something like that. This is a worked example of a single branching scenario where I show my actual prompts and results, with notes about what worked and what didn’t work. I’ll explain how I refined the prompts through the process and how I’m editing the draft results that ChatGPT creates.

Ideal path

In my last blog post, I used ChatGPT to draft two decision points. I usually start writing the “ideal path” first. That is, I write the version of the story where the learner picks all of the best choices. I did the same thing here, asking ChatGPT to keep writing from where I left off.

Prompt to continue on the ideal path (not a great prompt)

Continue the conversation starting with
You ask, “Rita, would you like to share your perspective first, or shall we start with Oliver?”

This prompt didn’t quite work the way I wanted because I didn’t give it enough information. It gave me a long conversation instead of just a little bit of the next part of the conversation before the next choice. However, I can use that draft ideal path and insert a few decision points.

Ideal path results from ChatGPT (with notes on where I’m adding breaks)

Rita takes a deep breath, appreciating the opportunity to express herself first.

RITA
(gently)
Thank you. I think I’ll go first. Oliver, I want you to understand that I value your enthusiasm and fresh ideas, but we have a proven process that has worked well for the team. It’s not about resisting change; it’s about building on what’s effective. Your dismissive attitude towards my experience makes me feel undervalued and unheard. I believe we can find a balance that benefits everyone.

Oliver looks taken aback, but he nods, showing that he’s willing to listen.

OLIVER
(trying to be receptive)
Okay, I hear you. I’m sorry if I came across as dismissive. I just thought my way could revolutionize things.

You, as the mediator, step in to guide the conversation.

(Insert decision point here)

MEDIATOR
(nodding)
Thank you, Rita, for sharing your perspective, and Oliver, for acknowledging how your approach might have been perceived. Now, Oliver, it’s your turn to share your thoughts. Please remember to be respectful and focus on the issue at hand.

OLIVER
(leaning in, earnestly)
I genuinely believe we can do better by adopting more modern methods. I mean, some of our processes are slow and outdated. I just want us to be more efficient and successful. But I get it; I shouldn’t have ignored the team’s experience.

MEDIATOR
(encouraging)
Thank you, Oliver. It’s clear that you’re passionate about improvement, and that’s valuable. Now, let’s work together to find common ground and explore how we can incorporate your ideas without discarding what’s already working well.

(Possible place to insert a decision point)

You guide the conversation, encouraging them to brainstorm potential solutions collaboratively. They begin to see each other’s viewpoints and recognize that a hybrid approach might be the best way forward.

RITA
(smiling slightly)
Perhaps we can test some of Oliver’s ideas on a smaller scale and see how they complement our current processes.

OLIVER
(enthusiastically)
Yes, and we can gather data to compare the results. If it’s successful, we can expand the implementation.

(Insert decision point here)

As the mediator, you help them outline an action plan with clear responsibilities and deadlines, ensuring they remain focused on their shared goals.

MEDIATOR
(optimistic)
That sounds like a great plan. Let’s start with a trial period and regular check-ins to assess progress. Remember, communication is vital throughout this process.

Rita and Oliver nod in agreement, visibly more open to collaboration and understanding each other’s strengths. The mediation session ends with a sense of optimism and a commitment to working together toward a more successful future.

First alternate path

Since my prompt for the ideal path gave me a response that was too long, I refined my prompt for the first alternate path. In this case, I was more explicit about how much content I wanted. I also combined this with a prompt for a question and options.

For this alternate path, I’m backing up to one of the earlier decisions about how to start the conversation.

Alternate path and decision prompt

Continue the conversation on an alternate path, showing what would happen if the user selected Option 1: “Alright, let’s get this over with. Rita, you start by stating your grievances.” This resulting conversation should have a negative result where the conflict escalates. Write dialogue for Rita, then Oliver, then Rita again. Then, write a multiple choice question with 3 options for how the user could continue the conversation at that point. Indicate which option is the best option that would deescalate the conflict.

ChatGPT-generated alternate path and choices

You decide to start the mediation session with a more direct approach.

MEDIATOR (bluntly) Alright, let’s get this over with. Rita, you start by stating your grievances.

Rita feels a bit put off by the abruptness but begins to express her concerns.

RITA (tense) Fine. Oliver, you come in here with your so-called “revolutionary” ideas and completely disregard the expertise and experience we’ve built over the years. It’s not just about the process; it’s about respect for what we’ve achieved together.

Oliver, feeling challenged, responds defensively.

OLIVER (defensive) Well, Rita, maybe if you were more open to change, we wouldn’t be stuck in our old ways. Your resistance to innovation is holding us back.

The tension escalates as their exchange grows increasingly heated.

RITA (angry) Holding us back? Your arrogance is holding us back, Oliver. You can’t just barge in here and throw everything we’ve built into chaos.

The conflict continues to escalate, and it’s clear that the mediation session has taken a negative turn.

MEDIATOR (intervening) Alright, let’s take a step back here. This isn’t productive. We need to find a way to communicate without attacking each other.

As the mediator, you try to steer the conversation in a more constructive direction.

Multiple Choice Question:

As the mediator, what do you do next to deescalate the conflict?

Option 1: Continue to press Oliver and Rita to express their grievances, hoping that they will eventually find common ground.

Option 2: Suggest taking a short break to cool off and reconvene when they’re both calmer.

Option 3 (Best Option): Acknowledge the intensity of the emotions and ask each of them to take a moment to reflect on what’s been said so far before continuing the conversation in a more respectful and solution-focused manner.

Map the structure in Twine

At this point, I decided I needed to start mapping out the structure in Twine. I need the visual to help me see the branching structure so I can see what I already have drafted and what I still need to create. I have my ideal path across the top, moving left to right, with the alternate paths below. If you haven’t tried Twine before, check out my previous posts on Twine.

Continue the alternate path

When I write branching scenarios, I always find it easiest to pick one path and then write it to the end. I’m following the same process here, even though I’m using ChatGPT for the drafts.

Alternate path and decision prompt 2

I used the same structure for my decision as before. I gave ChatGPT some guidance to make sure the best option uses the ground rules for mediation. If you were creating a branching scenario for a training project, this would be a good time to incorporate the information you gained from needs analysis and interviewing SMEs.

Continue the conversation, showing what would happen if the user selected Option 3 (Best Option): Acknowledge the intensity of the emotions and ask each of them to take a moment to reflect on what’s been said so far before continuing the conversation in a more respectful and solution-focused manner. This resulting conversation should have a positive result where the conflict deescalates and the mediation moves forward. Write dialogue for Oliver, then Rita. Then, write a multiple choice question with 3 options for how the user could continue the conversation at that point. Indicate which option is the best option that would follow the principles for mediation shared earlier in this conversation.

Alternate path from ChatGPT

As the mediator, you decide to acknowledge the intensity of the emotions in the room and suggest a pause for reflection.

MEDIATOR (calmly) I can see that this conversation has become quite intense. Let’s take a moment to collect our thoughts. Oliver, Rita, I’d like each of you to reflect on what’s been said so far and consider how we can move forward in a more respectful and solution-focused manner.

Oliver and Rita exchange glances, still clearly frustrated, but they nod reluctantly.

OLIVER (inhales deeply) Okay, I’ll take a moment to think.

Rita follows suit, taking a deep breath and visibly calming down.

RITA (sincerely) Alright, let’s try this again with a fresh perspective.

After a few moments of silence, Oliver begins to speak, his tone noticeably softer.

OLIVER (apologetic) Rita, I’m sorry for how I came across earlier. I didn’t mean to disregard your experience. I see now that I should have approached this differently.

Rita appreciates the change in tone and responds more positively.

RITA (appreciative) Thank you, Oliver. I also realize that I might have been resistant to change. Maybe we can find a way to combine our ideas for the benefit of the team.

The atmosphere in the room begins to shift towards a more cooperative and open dialogue.

MEDIATOR (encouraging) That’s a step in the right direction. Let’s work together to find a solution that leverages both your fresh perspectives and Rita’s experience. What do you both think would be the best way to move forward?

Multiple Choice Question:

As the mediator, what do you do next to facilitate a constructive conversation?

Option 1: Suggest that Oliver and Rita immediately start brainstorming solutions without further discussion.

Option 2: Ask Oliver and Rita to each share one positive aspect they see in each other’s ideas to build goodwill.

Option 3 (Best Option): Encourage Oliver and Rita to share their ideas and concerns, ensuring they actively listen to each other and fostering a collaborative approach to finding a resolution.

Rejoin the alternate path to the ending

This alternate path is now fairly positive, so I’d like it to rejoin the positive ending. I used a similar prompt to continue and create a final decision point in this path. Instead of saying who should speak first and second in the prompt, I changed it to “Write 2-3 lines of dialogue.” That keeps the length shorter. The result here sounds stiffer and less realistic, so it would need more editing. The best option here is to work together on an action plan–the same best option from my ideal path. That means I can have the learner rejoin the main path for the ending.

Decision point 3 prompt

Continue the conversation, showing what would happen if the user selected Option 3 (Best Option): Encourage Oliver and Rita to share their ideas and concerns. This resulting conversation should have a positive result where the two colleagues move forward toward a resolution. Write 2-3 lines of dialogue. Then, write a multiple choice question with 3 options for how the user could continue the conversation at that point. Indicate which option is the best option that would follow the principles for mediation shared earlier in this conversation.

ChatGPT response and choices

Encouraged by your guidance, Oliver and Rita start sharing their ideas and concerns openly.

OLIVER
(enthusiastically)
I think we could begin by implementing some of my ideas on a small scale, alongside our current processes, and then evaluate the results.

RITA
(appreciative)
That sounds reasonable. And we can involve the team in decision-making, so they feel valued and part of the changes.

Their conversation becomes more collaborative and solution-focused.

MEDIATOR
(optimistic)
Excellent. It’s clear that together, you can find a way to make positive changes that respect both your experience and Oliver’s fresh ideas. Let’s continue working on this plan together.

Multiple Choice Question:

As the mediator, what should you do next to facilitate a successful resolution?

Option 1: Suggest that they vote on whose ideas to implement, potentially leading to a winner-takes-all situation.

Option 2: Offer your own solution to end the conflict quickly and ensure both parties are satisfied.

Option 3 (Best Option): Continue to guide Oliver and Rita in collaboratively developing their action plan, ensuring they consider the best aspects of both approaches and how to integrate them effectively.

Rewrite an implausible choice

Option 1 provided doesn’t really make sense. There are only two employees, so a vote wouldn’t be a plausible action to take. Therefore, I asked for a replacement. This result is more usable.

Prompt for a replacement choice

Give me a replacement first option for this multiple choice question. This should be a wrong answer representing a mistake that a manager might make that violates one of the best practices for mediation. As the mediator, what should you do next to facilitate a successful resolution?

Option 1: Suggest that they vote on whose ideas to implement, potentially leading to a winner-takes-all situation.

ChatGPT’s replacement choice

Option 1 (Replacement): Take sides and express your own preference for one set of ideas over the other, hoping to influence the decision-making process.

Negative endings

Some of these negative choices should result in endings. In this case, I prompt both for the next part of the conversation plus the feedback.

Prompt for negative ending and feedback 1

Continue the conversation, showing what would happen if the user selected Option 1: Take sides and express your own preference for one set of ideas. This resulting conversation should have a negative result where the conflict escalates again and Rita and Oliver feel pitted against each other. Write 2-3 lines of dialogue. Then, write some feedback explaining why this choice did not follow the best practices for mediation explained earlier in this conversation.

ChatGPT’s negative ending 1

As the mediator, you decide to take sides and express your own preference for one set of ideas.

MEDIATOR (decisively) I believe Oliver’s ideas have more potential for success. Let’s go with those.

Oliver feels validated, while Rita becomes visibly upset.

OLIVER (triumphant) See, Rita? The mediator agrees with me.

RITA (angry) This is not fair at all. You’re supposed to be neutral!

Feedback:

Expressing a preference for one party’s ideas violates the fundamental principle of mediation, which is to remain neutral and impartial. By taking sides, you not only escalate the conflict but also erode trust and fairness in the mediation process. It’s crucial for a mediator to guide the conversation, not to dictate the outcome or show favoritism. Mediation should focus on fostering cooperation and finding mutually agreeable solutions.

Prompt for negative ending and feedback 2

Continue the conversation on an alternate path, showing what would happen if the user selected Option 2: Offer your own solution. This resulting conversation should have a negative result. Write 3-4 lines of dialogue. Then, write some feedback explaining why this choice did not follow the best practices for mediation explained earlier in this conversation.

ChatGPT’s negative ending 2

As the mediator, you decide to offer your own solution in an attempt to quickly resolve the conflict.

MEDIATOR (authoritatively) I’ve heard both your ideas, and here’s what I suggest: Let’s implement a combination of Oliver’s and Rita’s ideas. Oliver, you can take the lead on the initial phase, and Rita, you can provide guidance and support.

Oliver seems pleased with the proposed solution, but Rita appears frustrated.

OLIVER (content) Sounds fair to me. Thanks, Mediator.

RITA (disgruntled) This doesn’t address the core issue at all. We’re just sidestepping the real problem.

Feedback:

Offering your own solution without the full cooperation and agreement of both parties is not a recommended approach in mediation. It can lead to further resentment and dissatisfaction, as seen in Rita’s response. Effective mediation involves guiding the parties to reach their own mutually agreeable solutions, ensuring that both perspectives are considered and respected. The mediator should facilitate, not impose, the resolution.

Branching structure so far

Now, I have an ideal path (in green at the top) with a positive ending, plus an alternate path with a mix of good and bad decisions. I have two negative endings in the lower right. I still have several placeholders from the ideal path where I need to go back and add decisions plus alternate paths, and I need to flesh out some of the other choices and alternate paths. But, this is some clear progress in drafting this scenario.

Prototype and verify

As with everything generated by AI tools, anything you create with ChatGPT should be considered a draft that requires human review. These tools are great for coming up with ideas, but you’ll have to verify them with your SMEs and audience.

The bright side may be that tools like ChatGPT could help you generate initial drafts faster. That means it may be easier to get an initial prototype out for verification. You should always have a SME review the results for accuracy, especially for any specialized or high risk content (and honestly, you maybe shouldn’t be using these tools for those topics in the first place). But, if you can draft and prototype scenarios faster, maybe you can get it in front of a beta testing audience faster too. Giving yourself more time for testing and iteration seems like a valid approach to branching scenarios.

To be continued

I’ll continue expanding this series of posts as I flesh out more of the scenario. If you missed the prompts I used to generate ideas and start this scenario, check out How to Use AI to Write Scenarios.