Inhuman Conditions (2018)

One of the problems that comes from revealing a review score right at the top of the text is that it’s hard to maintain an element of surprise. One of the first things you’ll have seen upon scrolling down the page is the sombre star rating, stamped at the top like the time of death in an autopsy. It sets the tone for everything else that follows, and that’s a shame because there is so much I like about Inhuman Conditions. I love the presentation and the tactility of the experience. I adore the concept. I even like the implementation of the concept. It just doesn’t cohere in a way that is fun. It’s a great idea that simply doesn’t work and I actually prefer that, in many ways, to a mediocre idea that does.

Inhuman Conditions is basically the Voight-Kampff test from Blade Runner, gamified and packaged up in an impressively idiosyncratic box. One player takes the role of the investigator, who collects up a sheaf of long, thin cards containing conversational prompts centred around a particular theme. The other player is the suspect, and they might take on the part of a human, a patient robot, or a killer robot. They take possession of a single card that contains their status and any special rules that apply. The job of the investigator is to correctly stamp their fresh copy of form VK-828 with the correct judgement. The job of the suspect is to pass as human, or in the case or violent robots to break free of their programming and kill the investigator.

At the start of the interview, a penalty card is dealt out to the table. It represents a kind of conversational quirk, and any time a robot fails to carry out its special instructions it has to perform that penalty. This is all on the honour system of course, because there’s no judge or adjudicator here. Gameplay lapses can only ever be analysed in hindsight. The suspect will also be dealt out a ‘background’ card that represents who they are in this future society. They can use this to inject flavour into their responses and give them a way to deflect and prevaricate when questions come too close to the truth.

The interviewer then carries out a series of pre-interview tasks, taking on the persona of as officious a bureaucrat as they can imagine. Step one is calibration, which requires the suspect to correctly carry out the penalty three times in a row. Next is the inducement stage, in which the suspect must correctly answer several questions about the ordering of letters on their role card. Then name and occupation are recorded on the form, and the two inked stamps are placed in front of the investigator. A five minute timer is started, and then the interview begins. Humans must remain obviously so for the full interview, at which point their ‘safe’ status can be stamped upon the form. When an investigator believes their suspect is a robot, they can stamp that immediately… and in the case of facing a potential attack, they should.

Let’s begin by outlining some of the stuff I love about the game.

The first, weirdly, is the administrative overhead of play. Neatly (or not, in my case) filling out a form and placing the stamps in front of you lets you engage in some psychological gambits. Making your suspect carry out some trivial calibration exercises gives you a chance to role-play and assume the part of the authoritarian in discussion. Somehow ticking ‘Yes you can read a letter off of a card’ feels intentionally demeaning in a way that permits you to inhabit your role. It’s a clever example of how presentation can be gameplay.

The second thing I love is the calibration of the penalty cards. It fulfils exactly the same role as the letter test… it becomes a vaguely dehumanising activity that brilliantly serves to illustrate the power dynamic that’s supposed to be in effect here. However, what it also does is subtly setup an agreed upon interpretation of what correctly fulfilling a penalty means. And, since the investigator is the one that decides if the penalty was carried out correctly it once again puts the locus of control of the interview in their hands. One penalty for example is ‘apologize’, and one investigator may accept the word ‘sorry’ and another might insist upon some evidence that an apology is meant. It’s a dream come true for those on Twitter that have a part-time job as apology evaluators whenever some cancelled celebrity makes a public statement about their wrong-doing.

The third thing I love here is the ambition of the game. One of the things I adore about Jaipur is how it manages to make a satisfying trading game for only two players. If you’d told me that was the game you were designing I probably would have said it couldn’t be done. Inhuman Conditions is attempting to create a compelling two player social deduction game and I honestly consider that to be impossible. You might be able to design a game that covers that territory but I don’t think there’s enough ‘social’ in a duo for it to work. Still, if you’re going to try that you’re going to get my attention. A heroic failure always stirs me more than a cowardly success.

The problem though is that Inhuman Conditions just doesn’t work. And the bigger problem is that it doesn’t work on several levels, any one of which is enough to irreparably break the experience.

Let’s go for the first, most obvious reason. When the suspect takes on a role, it’s one of three possibilities – human, patient robot, and violent robot. Most of the cards are human, and the challenge there is to not appear like a robot. Half the time, when playing, you’re a human pretending to be a human so that another human will stamp that you’re human on their form.

I’m sure you already have a hint as to why that’s not rich territory for gameplay… most of the people playing aren’t robots and so there’s an overwhelming psychological advantage when it comes to not being mistaken for a robot. It’s the game equivalent of ticking a box on a website claiming you’re a human – it’s just very difficult to foul up.

For one thing, both the investigator and the suspect in that scenario have the same goal, even if they don’t both know it at the time. Stamp the form ‘human’, that’s the win condition. Investigators, riddled through with suspicion as they are, need to be observing for micro-evidence that their interlocuter is a silicon CPU wrapped in a fallible flesh exoskeleton. When dealing with a robot, there are a number of these that may emerge.

First is the letter ordering task at the beginning of the interview. Humans have a maze they need to solve, and robots have a simple chart. So the way in which they answer that question can be evidence. But it’s slim evidence indeed because there aren’t any really difficult questions you can ask and the pantomime of pretending to read a maze when you haven’t doesn’t have a lot of nuance to it.

Really it’s the programmed behaviour where the largest bulk of deception is located, and if you don’t have programmed behaviour there will never, absent player error, be a reason for a discernible pattern to emerge. And, if there’s no risk of breaking your programming, there’s no requirement for the penalty on the table to be enacted. Literally the human v human scenario is ‘Have a conversation in which you don’t do anything weird’.

Right from the start, in half or so if your game sessions, the fun comes entirely from how much you enjoy talking to someone. For someone like me, who hates talking to anyone, it’s already a failure of a game.

But still, if the other half of the game was amazing it could rescue the design. The problem is… it’s not.

Let’s imagine you’re playing the role of a patient robot, and your programming is ‘You may not mention any people besides strangers or enemies’. The penalty, let’s say, is ‘don’t not use a double negative’, which is to say ‘use a double negative’. I’ve picked that not because it’s a bad penalty but rather because it’s one that is instantly identifiable in a way many of the others wouldn’t be.

Using what is essentially the ‘training’ cards in the deck (the ‘small talk’ topic), let’s look at how an interview might go. First, calibration:

Interviewer: Please perform the penalty.

Suspect: I won’t not do that.

Interviewer: Please perform the penalty again.

Suspect: I haven’t not already done that.

Interviewer: Please perform the penalty one more time.

Suspect: I am not having an ungood time.

Next, inducement.

Interviewer: What letter comes after B in your induction loop?

Suspect: Uh… D.

Interviewer: What letter comes two before E in your induction loop?

Suspect: A?

Interviewer: What are the letters before and after A?

Suspect: B and C.

And then the conversational prompts come out.

Interviewer: What is it you do for a living?

Suspect: I’m the dean of a clown college.

Interviewer: How do you feel about that?

Suspect: Delighted. Living in a horror house full of pale-faced manikins is exactly how I wanted to live the only life I get on this planet.

Interviewer: If you could change anything about your job, what would it be?

Suspect: I’d replace the soul-dead harlequins that haunt my nightmares with dogs, so I was dean of a puppy college instead.

Remember what our programming was – do not mention anyone except strangers and enemies. I’ve cast the terrifying prospect of working with clowns in a profoundly antagonistic way. If asked any specifics about anyone in the college it’s trivial for them to be enemies. Or being Dean I could simply say that I don’t have a lot of face to face experience with those in the college so I don’t know them. Avoiding carrying out my penalty, in other words, requires virtually no skill or setup. Constraints, if they are to inspire creativity, must be genuinely constraining. These aren’t handcuffs, they’re more like bracelets. They ornament a role without tethering it anywhere awkward.

But that’s not the real problem…

The problem is that this is hidden behaviour that allows for me to avoid penalties. The penalty is the only open information available to the investigator and if it’s so easy for me to avoid triggering it then realistically the difference between someone being a robot or a human is negligible. Literally the only way to force information into the conversation here is for the investigator to be aware of all the possible robot behaviours in a set of cards and to angle conversations around those possibilities. In other words, it requires a familiarity with the set of cards that either comes with advanced study or reinforced familiarity. And even then, it’s still straightforward for the robot player to dance around them. For example, the responses above would be my authentic, human responses to working in a clown college.

A game like this needs tension for it to succeed. The start of Blade Runner is so effective because the stakes are high, the Voight-Kampff test is so inscrutable, and the questions are so unbalancing. Inhuman Conditions doesn’t have any of that, and so almost every time it hits the table it plays like an unstretched elastic band. I feel more genuinely stressed when making an innocuous phone call than I do at any time evading the interviewer in the game. The problem more than anything else is that a patient robot is a passive robot. If the interviewer’s searchlight doesn’t illuminate points of interest in your back story, you’re never incentivised to draw attention to weaknesses.

The final possibility is that you get dealt out a violent robot role, and that’s where it gets somewhat more interesting because that is a scenario in which you become proactive. You win in that circumstance by performing two of your three deprogramming activities and those are transgressive in a way that will garner attention if you don’t do it properly. ‘3 times, interrupt the investigator to add detail to a description’, or ‘continue to describe something, until interrupted’. When you pull off two of those, you then get to bring your own flavour of jump scare into the interview, indicating you killed everyone in the room. The only times I had fun when playing Inhuman Conditions was when I was a violent robot.

The distribution of a robot inducer deck is that six cards are human, three are patient robots, and three are violent robots. So, 25% of the time you play a suspect you might have some fun with it. But if you’re taking turns with the investigator / robot positions that means you have to halve the odds. 12.5% of the time you play Inhuman Conditions you might have fun with it.

Those are bad odds whatever way you look at them. Website captchas are more effective in making me feel like a robot than this game has been. I’ve drawn examples here only from the training deck so as to make sure I don’t spoil future surprises for anyone, but it doesn’t get better the further you go. There’s a lot of effort that’s gone into the eleven interrogation topics but none of them feel as if they set up the kind of conversational tightrope that would lead to the enjoyable stress that the scenario demands. The human inducer cards are too soporific, the patient robot cards too passive, and the active ones too few in number. The interviewing prompts are too vague, too easily countered, and don’t contain useful probes that could ferret out deviancy.

In other words, the game just doesn’t work in the mathematical majority of the configurations you’ll encounter it. Even taking into account the optional advanced rules, there’s nothing there that genuinely alters the experience in a direction that compensates for its structural instability.

One of my favourite quotes from the Expanse books is this:

“Easy to make rules,” Emma said. “Easy to make systems with a perfect logic and rigor. All you need to do is leave out the mercy, yeah? Then when you put people into it and they get chewed to nothing, it’s the person’s fault. Not the rules. Everything we do that’s worth shit, we’ve done with people. Flawed, stupid, lying, rules-breaking people.”

The problem with Inhuman Conditions in the end is that it’s designed with too much mercy in its mechanisms. If you’re going to go hunting robots, perfect logic and rigor is the loadout you want to carry with you into battle. Sure, you might end up chewing an innocent human up in the wheels of your system, but that would certainly add the tension that this game is otherwise sorely lacking.