Artificial Condition by Martha Wells

One of the things that science fiction can do better than many other genres of literature is to take an abstract philosophical or metaphorical problem and make it very, very literal. “Am I forever defined by my past?” is a popular introspective question. “How do I deal with all of these other beings around me?” is another common question. It’s fair to say that no one has tried to answer them as a self-aware, human-machine hybrid that’s programmed to protect humans but has overridden its governance module and, in addition, went on a killing spree some time back. Well, tried to answer them from that perspective outside of fiction at any rate.

That’s about the shortest possible background of Murderbot, the first-person narrator of Artificial Condition, the second of (to date) four novellas collectively titled The Murderbot Diaries. I had not read the first one, All Systems Red (Doreen’s review is here), when I picked up the second as part of this year’s Hugo voting process. Wells provides enough of Murderbot’s background to piece things together, but it definitely would be better to begin at the beginning.

At the beginning of Artificial Condition, Murderbot is trying to be inconspicuous, and to book passage to the system where it went out of control. The two goals appear contradictory, even though Murderbot has relinquished its armor and opaque visor, until it connects with a bot-driven transport headed where it wants to go. Hilarity doesn’t exactly ensue, but there’s a bleak humor that Marvin would definitely recognize.

When constructs were first developed, they were originally supposed to have a pre-sentient level of intelligence, like the dumber variety of bot. But you can’t put something as dumb as a hauler bot in charge of security for anything without spending even more money for expensive company-employed human supervisors. So they made us smarter. The anxiety and depression were side effects. (pp. 10–11)

Murderbot is looking forward to 21 cycles (analogous to days) of uninterrupted media consumption while traveling. It hasn’t reckoned with the transport ship suddenly intruding and saying “You were lucky.”

I sat up. It was so unexpected, I had an adrenaline release from my organic parts.
Transports don’t talk in words, even through the feed. They use images and strings of data to alert you to problems, but they’re not designed for conversation. I was okay with that, because I wasn’t designed for conversation, either. …
I poked cautiously through the feed, wondering if I’d been fooled. I had the ability to scan, but … with all the shielding and equipment around me I couldn’t pick up anything but background readings from the ship’s systems. … But the presence in the feed was too big and diffuse for a human or augmented human, I could tell that much even through the feed walls protecting it. And it sounded like a bot. When humans speak in the feed, they have to subvocalize and their mental voice tends to sound like their physical voice. Even augmented humans with full interfaces do it.
Maybe it was trying to be friendly and was just awkward at communicating. I said aloud, ‘Why am I lucky?’
‘That no one realized what you were.’
That was less than reassuring. I said, cautiously, ‘What do you think I am?’
It said, You’re a rogue SecUnit, a bot/human construct, with a scrambled governor module. It poked me through the feed and I flinched. It said, Do not attempt to hack my systems, and for .00001 of a second it dropped its wall.
“It was enough time for me to get a vivid image of what I was dealing with. Part of its function was extragalactic astronomic analysis and now all that processing power sat idle while it hauled cargo, waiting for its next mission. It could have squashed me like a bug through the feed, pushed through my wall and other defenses and stripped my memory. Probably while also plotting its wormhole jump, estimating the nutrition needs of a full crew complement for the next 66,000 hours, performing multiple neural surgeries in the medical suite, and beating the captain at tavla. I had never directly interacted with anything this powerful before.” (pp. 11–12)

“Awkward at communicating” describes both the transport and Murderbot very well. It’s interesting to see these two intelligences reasoning through something that neurotypical humans generally take for granted. In due course, they figure out that neither is going to do the other grievous harm, and that there are things that they could enjoy doing together, for very bot-like values of “enjoy.”

The ship asks Murderbot to go back to watching a particular show: “It was called Worldhoppers, and was about freelance explorers who extended the wormhole and ring networks into uninhabited star systems. It looked very unrealistic and inaccurate, which was exactly what I liked.” (p. 14) Murderbot is cranky and initially refuses.

Two minutes later it repeated the ping and the request.
I said, ‘Watch it yourself.’
I tried. I can process the media more easily through your filter.
That made me stop. I didn’t understand the problem.
It explained, When my crew plays media, I can’t process the context. Human interactions and environments outside my hull are largely unfamiliar.
Now I understood. It needed to read my reactions to the show to really understand what was happening. Humans used the feed in different ways than bots (and constructs) so when its crew played their media, their reactions didn’t become part of the data. (p. 15)

The ship’s reaction puzzles Murderbot even more.

It didn’t complain about the lack of realism. After three episodes, it got agitated whenever a minor character was killed. When a major character died in the twentieth episode I had to pause seven minutes while it sat there in the feed doing the bot equivalent of staring at a wall, pretending that it had to run diagnostics. Then four episodes later the character came back to life and it was so relieved we had to watch that episode three times before it would go on.
At the climax of one of the main story lines, the plot suggested the ship [in the story] might be catastrophically damaged and members of the crew killed or injured, and the transport was afraid to watch it. (That’s obviously not how it phrased it, but yeah, it was afraid to watch it.) I was feeling a lot more charitable toward it by that point so was willing to let it ease into the episode by watching one to two minutes at a time.
After it was over, it just sat there, not even pretending to do diagnostics. It sat there for a full ten minutes, which is a lot of processing time for a bot that sophisticated. Then it said, Again, please.” (pp. 15–16)

None of this is the main story of Artificial Condition, which concerns some human researchers trying to recover key data from an employer who has decided to screw them over (efforts that go bad in ways that are entirely predictable for Murderbot), but I think it gets to the heart of what the novella is about: beings that feel normal to themselves, but appear very peculiar from a typical external human perspective, trying to figure out how to deal with the other beings they come in contact with. Beyond that, Murderbot wants to find out how thoroughly it has earned its nickname. Over the course of Artificial Condition — which really is a lot of fun to read for its main story — they both get the only kind of answers anyone does for those questions: partial and provisional.