[ FEATURE REQUEST ] - Say something

[ElsenIsCute]

Overview:

I had an idea for a simple dialog. Basically, you'd ask Natsuki if you want to type her something and if she's on normal or higher she'll say yes.

Must haves:

After a short intro dialog, if she says yes a dialog box would pop up where you can type something to say, and when you press enter she'd have some kind of generic response to it. This would allow you to comment your thoughts after Natsuki has a random topic, and so that she isn't just talking mindlessly into a void, with barely any initiative from the player.

Could haves:

Detection of certain keywords such as "I love you". Maybe a mirror topic like "Can I vent?". Another topic of asking Natsuki a yes or no question and it'd work like an 8 ball, thus adding to the conversational aspect more.

[Blizzardsev]

I think the problem here ultimately is execution once again, and you've highlighted that dreaded word we always try to avoid being, as mentioned in #662 and #717 : generic!

Something important to understand about direct user input is that processing it is hard - and getting the correct context from plain text is even more challenging. Let's go with the idea of keywords as you highlighted in your suggestion:

• We take in the user's input: let's say the player has just finished listening to Natsuki's topic about her art classes, and responds:
  • I love art, hate class but I actually draw all the time Natsuki !
• We try and identify keywords. Entire libraries already exist for this, and no implementation is perfect (AI is another story altogether, so we'll leave that out of this discussion): let's assume we get the following keywords from that dialogue:
  • ("ART", "HATE", "CLASS", "DRAW", "NATSUKI")
• How do we process the keywords?
  • Do we match them to a topic, and have a response based off that? We'd be looking at a response per topic, which is just not realistic imo.
  • Do we just look at the keywords to keep things simple? How do we determine how Natsuki should respond? Should each keyword have a positive/negative attribute, with an overall score determining if Nat is happy/upset/normal at the result?
  • Do we have a map of keywords to responses? That's going to be a very long list!
  • What conclusions can we draw from the keywords?
    • The player hates drawing in class?
    • The player wants to draw Natsuki?
    • The player's class hates art?
    • The player hates art, and their class all drew Natsuki? (and so on)
• What is Natsuki's response?
  • If we keep things simple (let's say a generic response based on if Nat liked or disliked the overall "score" of the keywords), then we end up with some really strange (and not satisfying responses!) to the questions!
    • "I know, right? Ehehe. I'm just glad you agree, player!" : Agree with what? Nat doesn't say she hated art class!
    • "Yep! Couldn't have said it better myself!" : This makes little sense - Natsuki just admitted she can't draw!

This is just one scenario, but hopefully it illustrates some of the challenges a feature like this would face.

Adding to that list:

• False positives: The player vents about something and swears, perhaps combining profanity with Natsuki's name in the keywords: Nat could easily come to the conclusion they're saying nasty things about her, and react accordingly!
• False negatives: How many ways are there to tell Natsuki you love her?
  • "<333 u Natsuki!"
  • "Love you bunny!"
  • "Love you sm"
  • "Heart you NatNat"
• Precedence: As a concept, this was actually how Monika After Story (MAS) designed their player interaction to begin with! We know that this solution isn't a great way of handling input from the user because we've seen it tried - and retired - before!
• Maintenance: Something like this immediately adds dependencies: if we avoid being generic, then we chain ourselves to having to update this every time we look to add something, and keep it in mind whenever we add a new feature. If we are generic, then I've already covered my thoughts here.
• User experience: Have you ever asked Siri/Google/Alexa a question?
  • "Sorry, player... I'm... kinda drawing a blank here."

I can see the potential here, and more interaction between Natsuki and the player is a good thing: but at the end of the day I just can't see this being an approach we'd be interested in given the above.

@RaionArt @Obsidian-Knife thoughts here?

[Blizzardsev]

I should add that I do remember you mentioned previously in our community server that this is something you'd be interested in working on: and by all means, you're welcome to do so with your own installation!

[RaionArt]

I feel like a request like that would be more of a new mod itself. Some programs like ChatGPT or other AI's simply exist to simulate a dialogue like that. On one hand to actual code a chat bot would be already too much to ask for imo, but additionally on the other hand it destroys the reason JN exists in the first place. We enjoy working on JN - to think about how she would react in certain scenarios or what her opinion about xy might be, etc. This would be kinda pointless if she would just answer with generic sentences.

I'm sure this could be an additional for a potentional submod if people feel like working on it and once we are able to provide support for it, but I don't see it as an addition for the official JN.

[Blizzardsev]

Closing this one as per our explanation. There's no real way to implement something like this to our (or our players) satisfaction, at least in a reasonable timescale and without giving ourselves plenty of headaches later.