NPC Dialogue Trees

Beep beep boop

The simplest way video games can fake interactivity when it comes to character interactions is through a dialogue tree. You allow the player a very limited repertoire of things to say or questions to ask, and give non-player characters (NPCs) a set of canned responses they can offer in reply. It was and remains an effective parlor trick because it requires no processing power, and most interactions with NPCs tend to be brief and perfunctory. An example of what a tree could look like is this:

I’m sorry I’m writing about

Simon Laird

again, but the guy provides a cornucopia of so many things wrong with The Discourse. Simon responded to my post calling out how he uses obfuscatory language to provide cover for positions that are embarrassing to be angry about when otherwise broadcast transparently.

The example I outlined was crystal clear: Simon turned “The Dems subpoenaed records from VDARE specifically over the course of a resource misallocation investigation” and obfuscated it into “The Dems want to shut down independent media!” Simon neither challenges that he obfuscated, nor does he offer an alternate theory for why he avoided speaking transparently. So my primary thesis stands.

But beyond that, Simon goes on an absolute tear accusing me of all sorts of things: I’m racist, evil, a psychopath, blah blah blah. His crown-jewel criticism is that I’m a “vicious aggressive tribalist”:

The gist of Yassine Meshkout’s post (and the gist of our substack Notes exchange) is that: i. Yassine presents examples of extreme, illiberal, authoritarian action being taken against people he doesn’t like. ii. At some level, Yassine realizes that those actions starkly violate the liberal principles he claims to hold. iii. Then Yassine lies or lies by omission, OR invents a bunch of ludicrous, ad hoc, arbitrary reasons why the illiberal state aggression against people he dislikes is actually like totally ok and it actually doesn’t contradict liberal values because of some obscure bullshit reason that Yassine made up just now for this specific circumstance, and only people as sophisticated as Yassine are sophisticated enough to understand this bullshit reason, and it’s actually like so nuanced and complex so who even knows anyways, and if you believe in equality under the law then you’re like totally uptight, man. I think we’ve all encountered people like Yassine. Let’s speak frankly about this. THE MOST PROBABLE EXPLANATION OF WHY PEOPLE ACT LIKE YASSINE IS BECAUSE THEY ARE VICIOUS, AGGRESSIVE TRIBALISTS. They are acting out of the deep evolved instinct to override their conscience when there is an opportunity to brutalize the other tribe.

Simon is describing a very real phenomenon! One he’s already experientially familiar with given his own emotionally unhinged conduct. If Simon was describing my conduct accurately — that I align myself with a “tribe”, pretend to hold principles, but make up excuses to oppress the out-group — then I absolutely would deserve condemnation.

The problem here is that he’s robotically following a primitive NPC dialogue tree without checking if it actually applies. If Simon the NPC was speaking to another partisan NPC from the other tribe, then this particular angle of attack would actually work! There’s absolutely no shortage of shills on either side of the aisle willing to make up new standards whenever convenient for their tribe. Republican Senator Lindsey Graham famously taunted “I want you to use my words against me” in 2016 over how it’s inappropriate to fill a SCOTUS vacancy during an election year, before pulling a 180-degree handbrake turn 4 years later. Now that the Dems are no longer in power, they suddenly want to hold on to the filibuster, with Democrat Senator Brian Schatz saying “I’m going to try not to make a mess of my position on this one”.

So am I being a shill? Investigating oneself isn’t ideal and so I welcome harsh feedback if I falter on this endeavor. The first axis the shill criticism fails is that Simon misidentifies my “tribe” as being the Democrats. I’ve been a boring bog-standard George Mason University libertarian for almost 20 years now. When Obama was in power, I criticized his appetite for drone strikes and the Left’s sudden aversion to protesting war. When Hillary Clinton lost to Trump, I was thrilled because she represented to me an elitist form of political entitlement, and I applied to work for Trump’s administration! Nowadays I clearly and demonstrably am much more aligned with the liberal establishment, but only to the extent they offer sanity on issues I care about. I would be absolutely thrilled if the Republican party stopped being the bastion of conspiracy nuts and didn’t make shitting on free trade its crowning obsession.

The second axis the shill criticism fails would have been dispelled by a brief perusal of my archive. I care about criminal justice abuses, and even though I believe the J6 defendants are a bunch of dumbasses, I nevertheless criticized how many of them were reflexively denied pre-trial release, and I lambasted the prosecution for hiding evidence in the Nordean trial. I care about gun rights, and so I praised the snare trap that Clarence Thomas and the Trump appointees laid out for gun control liberals to step into. I care about the right to self-defense, and even though I hold sympathies with BLM protestors on some policy issues, I immediately described Kyle Rittenhouse’s conduct as unambiguously justifiable. I care about abuses of the legal system, and so I criticized the defamation bear trap Trump unjustly found himself in for denying a rape accusation, and excoriated the generous accommodations Seattle officials got for destroying evidence. I’ve also repeatedly pointed out the delusional hypocrisy some leftists hold about anti-white racism being legally permissible.

I don’t know about you, but I’m noticing a distinct pattern where I actually care about the issues I claim to care about, regardless of how I might feel about the individuals affected. I could go on and on and on, how much more do you need?

Simon didn’t bother a peek and instead followed an embarrassingly primitive dialogue tree of if anyone disagrees with me > assume they’re liberal > assume they’re a shill and then rummaging through the drawer of meme talking points to respond with. That explains why he just assumed I must be as unprincipled as he is and would not have wanted Hillary Clinton to be prosecuted for her private email server shenanigans. This is a stance absolutely divorced from reality given how many times over the years I’ve warned against the generous treatment the politically-connected receive compared to normal people.¹ Here’s me saying it again: Hell yes Hillary Clinton should have been prosecuted! As much as I might rail against over-incarceration in general, we definitely have an under-incarceration problem when it comes to politicians.

This NPC dialogue tree tactic extended to a hilarious comment to Simon’s post by

Eugine Nier

where he begrudgingly acknowledges I’m an atheist but nevertheless claims I “appear to be a cultural Islamist”:

Hahahahahahahahah

What in the fuck could this possibly mean?? Presumably it’s a dialogue tree along the lines of Yassine is an Arab > Arabs are generally Muslim > Islam is bad > ergo Yassine is bad for reasons I’m incapable of articulating.

The problem with relying on an NPC dialogue tree is that it doesn’t allow for flexibility when encountering novel circumstances. If you create a druid NPC in your video game whose sole purpose is to offer players a quest to return wolf pelts from the forest, that druid will be incapable of riffing on the current geopolitical crisis in the nearby kingdom, at least not until LLMs enter the scene. If I was a Muslim, Eugine would presumably have a bevy of valid criticisms to levy against me, much of which I (meaning real atheist me) would agree with! But because we’re off-script, all we can hope for is his glitched-out cognitive capacity collapse.

---

The current media ecosystem heavily encourages the dissemination of knowledge through inchoate short-burst snippets. What you would hope would warrant baseline skepticism somehow gets circumvented, and “saw it in a tweet once” becomes an acceptable method of citation. I’m not prone to being an alarmist, but I see this trend as a serious and foundational threat to our modern liberal order and civilization in general. I don’t know what to do about it except unplug and hope for the sea to change.

I’ve recently deleted several social media accounts, including my Twitter. Part of is the undeniable realization that it’s all a giant waste of time, but that wasn’t always the case! I didn’t really get into the groove of things until early 2022 when I finally connected with the right people, and it was cool and fun and chill to directly interact with a host of folks whose work I’ve long admired. The quality of engagement has steadily downgraded after Elon Musk took over, especially after he changed the (admittedly needlessly cloistered) old Blue Check to be a pay-to-boost feature instead. Suddenly you got a bunch of inane slop surfacing to the top of the algorithm. The real death knell though was how perfect of an environment this was to inject chatGPT bots into the soup.

Benn Jordan, a synth producer of all people, put together a fascinating and disturbing video about how serious of a issue the bot pollution problem is on Twitter. Hank Green also raised the same point about the rising utilization of AIs spewing discord-sowing babble. It’s already very difficult to tell who is or isn’t real on the internet but I have absolutely no idea what platforms can do to stop the inevitable end point of a slush pit of AIs talking to other AIs.

Even if that calamity is somehow averted, we already have human NPCs robotically working off talking point scripts, and that works OK so long as they’re up against other human NPCs robotically working off their own talking point scripts (and they often are). Simon would have been far better served generating his reply to me via chatGPT because it would’ve been trivial to just copy/paste a handful of my posts with the prompt “write me a savage takedown for this dipshit who hurt my feelings”. The end product might still have been retarded, but it wouldn’t have been as transparently retarded.

As LLMs continue to moon towards higher capability echelons, and as human NPCs realize the benefits of augmenting their scripts with LLMs, the problem is just going to get worse. I wish I had more optimism, but every signs points to us expecting more and more of this simpleton discourse tennis.

1. It also probably explains the oddly repeated references to how if I was raised in Russia I would have been an apologist for Navalny’s murder, presumably that’s the if lib shill > sympathetic to Navalny dialogue branch of the tree. Who knows.