Newcomers

It's a useful regex blunderbuss approach when the AI Engine won't behave. But they can get a LOT more complicated than that, believe me!

MickMcA: The 'dev' level doesn't just reflect the number of lines in your bot, and some things (such as adding a keyphrase with only one response) can actually LOWER your dev level. The exact formula is known only to the Prof.

That's a myth, easily verifiable. What happens when you add a keyphrase with only one response is that it'll only add a small fraction of a point to the Development score. What'll lower it is AI scripts and possibly seeks (I haven't been able to nail down whether the seeks do it).

I suspect it has something to do with these features being newer additions to the engine (well, relatively newer) and maybe the Dev score wasn't considered at the time.

Except for seeks with only one response, I've found that AIScript and Seeks raise the development. Perhaps there's an element of randomness in the Professor's secret formula.

The puzzle is worse than I expected.

Let me offer a possiblity to consider. There are two ways to program a bot, what I call brute force (anticipate every KeyPhrase, at its most "primitive") and what I would call "elegant" (guess where my sympathies lie...), which attempts to build a rule that captures as many true positives as possible. Is it possible that the Dev score is weighted toward brute force?

Just thinking out loud.
M
BTW: Pocque has been mainly built by brute force, and my other bot, with the same number of lines, is heavily weighted the other way, but rated at 1/3 of Pocque's "development."

Two questions:

1. Is there some way to turn off the "volunteer" responses? I have a character who is explicitly primed [50,0] to say what their favorite movie is. Instead, I get random answers that are totally out of character, courtesy of some response data in the background. I finally got the scripted response by jacking the Rating up to 150.

2. [related] Where do the bizarre substitutions come from? I asked a character if they had read For Whom the Bell Tolls, and they responded, "No. Is it fun to victimize For Whom the Bell Tolls?" What?? I understand that the bots need to have access to more than just our responses, to seem spontaneous, but what's the point of building a bot around a character if the engine is going to ignore the "in character" stuff that's sitting there waiting to be used?

I have had this problem with my "perv trap" as well, and I can't seem to defeat it. With a hundred traps set up, the bot still responds to sexual invitations with comments like, "That would be a fun way to spend the weekend," etc., which is not only totally out of character, but gives the sex boy a taste of exactly what he wanted.

well, you * read * For Whom the Bell Tolls rank=30 works fine for me. No need for regexes, raw mode or blunderbusses.

I found one cause for the (read=victimize) problem. I had used this approach (tildes are tabs):
Do you like to (verb)?
-(verb)? I don't know how to (verb).

I thought that would repeat the identifed verb. Apparently it substitutes random verbs instead. Revising as
~(key1)? I don't know how to (key1).

I've done that by mistake on a few phrases (put in (adj) or (verb) instead of (key#)). So it's easy to do even when you know it should be a "key".

I'm going to start experimenting with the sentence parsing tools in the PF2 addenda. I get into cases where I don't want (key1) unless it is the verb.

i know this is of topic and sounds stupid but can i add some of you guys to my friends list? you can say no if you want

No, it's perfectly within the topic bounds.
Yes, you can add anyone you like. Or did you mean 'may'? Can't speak for anyone else, but you're welcome to add me.
We don't even have a way of knowing - why would we need/want to say no?