When Robot and Crow Saved East St. Louis

A new short story about a disease surveillance robot whose social programming gets put to the test.

A drone with eyes flies next to a crow over the skyline of East St. Louis.
Lisa Larson-Walker

Each month in 2018, Future Tense Fiction—a series of short stories from Future Tense and ASU’s Center for Science and the Imagination about how technology and science will change our lives—will publish a story on a theme. The theme for October–December 2018: Work.

It was time to start the weekly circuit. Robot leapt vertically into the air from its perch atop the History Museum in Forest Park, rotors humming and limbs withdrawn into the smooth oval of its chassis. From a distance, it was a pale blue flying egg, slightly scuffed, with a propeller beanie on top. Two animated eyes glowed from the front end of its smooth carapace like emotive headlights. When it landed, all four legs and head extended from portals in its protective shell, the drone was more like a strangely symmetrical poodle or a cartoon turtle. Mounted on an actuator, its full face was revealed, headlight eyes situated above a short, soft snout whose purple mouth was built for smiling, grimacing, and a range of other, more subtle expressions.

The Centers for Disease Control team back in Atlanta designed Robot to be cute, to earn people’s trust immediately. To catch epidemics before they started, Robot flew from building to building, talking to people about how they felt. Nobody wanted to chat with an ugly box. Robot behaved like a cheery little buddy, checking for sick people. That’s how Robot’s admin Bey taught Robot to say it: “Checking for sick people.” Bey’s job was to program Robot with the social skills necessary to avoid calling it health surveillance.

Robot liked to start with the Loop. Maybe “like” was the wrong word. It was an urge that came from Robot’s mapping system, which webbed the St. Louis metropolitan area in a grid where 0,0 was at Center and Washington. The intersection was nested at the center of the U-shaped streets that local humans called the Loop. A gated community next to Washington University, the Loop was full of smart mansions and autonomous cars that pinged Robot listlessly. Though it was late summer, Robot was on high alert for infectious disease outbreaks. Flu season got longer every year, especially in high-density sprawls like St. Louis, where so many people spread their tiny airborne globs of viruses.

Flying in low, Robot followed the curving streets, glancing into windows to track how many humans were eating dinner and whether that number matched previous scans. Wild rabbits dashed across lawns and fireflies signaled to their mates using pheromones and photons. Robot chose a doorway at random, initiating a face-to-face check with humans. In this neighborhood, they were used to it.

A human opened the service window. The subject had long, straight hair and skin the color of a peeled peanut.

“Hello. I am your friendly neighborhood flu fighter! Please cough into this tissue and hold it up to the scanner please!” Robot hovered at eye-level, reached into its ventral service trunk, and withdrew a sterile sheet with a gripper. This action earned a smile. Robot smiled back, stretching its dog-turtle mouth and plumping its cheeks. Humans valued nonverbal emotional communication, and it was programmed with an entire repertoire of simple exchanges:

If human is angry, then Robot is sad.

If human is rude, then Robot is embarrassed.

If human is happy, then Robot is happy.

The human coughed and Robot did a quick metagenomic scan, flagging key viral and bacterial DNA before uploading sequence data to the cloud. Other bots would run the results against a library of known infectious diseases and alert the CDC if any were on the year’s rolling list.

Six days later, Robot headed across the Mississippi River to East St. Louis. Here, heat and rain had eroded the pavement until its surface was as pocked and fissured as human skin. The first time Robot performed health surveillance in this area, nothing fit its generic social programming. Buildings marked as unoccupied were clearly full of humans. Occupant records did not match the names and faces of occupants. People spoke with languages and words that did not match known databases. As a result, Robot could not gather adequate data. When Robot requested help with this problem, Bey was the only CDC admin who responded. She communicated with Robot from Atlanta via cellular network, using audio.

“Not all humans behave or speak the same way,” she told Robot. “But you can learn to talk to anyone. Gather data. Extrapolate from context. Use this.” And she sent Robot a blob of code for natural language acquisition and translation. Very quickly, Robot learned that humans used slang, dialects, sociolects, and undocumented lexicons. Bey also sent several data sets taken from an urban studies lab, which supplemented Robot’s map data. It turned out that not all humans lived in the same domicile for two years on average; not all residences had cars and rabbits outside. Some humans lived in places that were not tagged as domestic spaces. Some humans did not use government-assigned identifiers. But all of them could get sick.

There was a small neighborhood of soft textile homes underneath the freeway. It did not exist on official maps. Robot knew it because of Bey’s algorithms.

“Hello!” Robot said, landing on the porch of a blue fabric house. It spoke a dialect that was popular here. “I am checking to make sure you are healthy! Please say hello!”

A human rustled inside, then unzipped the door.

“Hi Robot.” The human had brown eyes and facial symmetry that matched previous records. It was the same human as last month.

“Please cough into this tissue and allow me to scan.”

The human smiled, and Robot knew why. The word for cough in this dialect was a pun for something the humans found endlessly amusing. There was a more formal word for cough, but compliance was higher if Robot used the pun. Higher compliance rates meant better data.

“Robot, I think my friend Shareeka is sick. Can you please check on her?” The human was worried, and Robot responded with a sad/concerned expression.

“Where is Shareeka?”

“She’s in the new building on State near 14th? On the upper floors that aren’t finished. I bet you could fly right in.”

“Thank you for your help.”

The human petted Robot’s head. It was the most common form of physical affection that Robot had documented in its four years and eight months in the St. Louis metropolitan area.

Protocol held that Robot should follow up on disease reports immediately, so it flew to the new building on State. Like the textile neighborhood, this building was not a designated residential area. It was a gray box on Robot’s official map. But visual sensors showed a reflective spire, with 20 floors wrapped in steel and glass. Five floors rose like a skeletal crown on top, exposing its steel beams, pipes, and drywall. Coming from inside were the sounds of human life: music, conversations in six languages, babies crying, food sizzling on hot plates. Robot could see electricity cascading down wires from solar panels bolted to the outside of windows. Residents tuned the data network with satellite dishes made from woks and metal cans. From Robot’s perspective, it was exactly like other residential buildings with a few cosmetic differences.

Extending its feet and head, Robot landed on the lowest open floor, then walked to the interior, asking for Shareeka. A juvenile human opened a green door and said hello. The human had short hair, woven into pink extensions, and a well-worn text reader in one hand.

“Hello! I am Robot, and I want to make sure you are healthy. A nice person told me that Shareeka might be sick. Can I meet Shareeka?” Robot used the same dialect it had in the fabric neighborhood, adding enhancement words that signaled benevolence.

The human made a neck motion that meant “no.”

“I am a friend who only cares about whether you are well. I am worried about Shareeka.” Robot made a sad face.

The human made a sad face too. “Shareeka left a couple of days ago. I don’t know where she is.”

“How do you feel today?”

“I’m kind of stressed out about school,” the human said. “How are you feeling?”

It was very rare for a human to ask Robot how it felt, and there was no stock answer or expression available. So Robot answered as literally as possible. “I am not sick because I am a machine. But I am worried that you are sick. Would you cough into this tissue and allow me to scan it?”

“Are you going to sequence the DNA right now?” The human was intrigued.

“Yes! But I will work with bots on the data network to figure out if anything dangerous is in there.”

“I know. You have a list of known infectious diseases and you’ll search for a match. We learned about it in biology class.” The human smiled, and Robot smiled back.

“Yes! That is what I will do.” It held out the tissue.

The human coughed on it and studied Robot very carefully as it conducted the scan.

“How do you make sure that you don’t mistake somebody else’s microbiome for mine? Do you sterilize your hand every time?”

“Yes I do.” Robot uploaded its data and talked at the same time. “What is your name?”

“Everybody calls me Jalebi.”

“You are named after a fried, spiral-shaped sweet soaked in sugar water.” Humans enjoyed it when Robot recognized the meaning behind their names.

Jalebi nodded. “When I was a kid, I ate so many that I passed out. Too much sugar. So my brother started calling me Jalebi.”

Robot was having difficulty making a connection to the cloud. “I am going to go back outside to talk to the network. It was nice to meet you Jalebi.”

“Wait—what’s your name?”


“That’s your name? I thought that was your … race.” Jalebi used an ambiguous word that could also mean “species.”

“It’s my name,” Robot replied.

Robot stood in the darkness beneath the moon, above the neighborhood lights, in the unfinished hallway open to the air, and called for the cloud. There was nothing. It called for Bey. There was no answer. It sent an emergency email to the CDC surveillance team list and got an error message. It called and called, charging up every morning in the sunlight and powering down at midnight. After seven days, it got a text message from an unknown private number:

Hi Robot. It’s Bey. I can’t be your admin anymore. I’m really sorry because it was nice to know you. Unfortunately the CDC lost its funding. I work at Amazon Health now, but we aren’t allowed to network with open drones like you. I don’t think anyone is going to shut you down or collect you, so I guess you can do whatever you want. If anything really bad happens, text me here on my private number. I hope the language acquisition algorithm is still helping!

For the first time, Robot made a sad face that nobody could see. It wasn’t sure what “really bad” meant, but its models of human communication suggested that Bey referred to an outbreak. The problem was that Robot had no way to conduct a typical surveillance circuit without somewhere to upload its data for analysis. Plus, it was going to run out of sterile tissues. That’s what happened last year when the government shut down and Walgreens froze its CDC account. Robot used the government shutdown scenario to model its current situation, and predicted that it meant the Walgreens account would be frozen for an indeterminate length of time. The 5,346 sterile tissues remaining in its chassis were the last it would ever have. The sterilizing gel for its gripper was already running low.

Bey said Robot could do whatever it wanted, which was the kind of thing humans said when they expected it to predict which data-gathering task should be prioritized. Based on current supply levels and its onboard analysis capabilities, Robot determined it should focus on learning local languages and human social habitation practices. It would attempt to reach the cloud every morning, and would reprioritize if disease analysis systems became available again. Robot thrust its head out of the pocked oval of its body, a determined smile on its face. In the absence of a human, the expression was intended only for a theoretical model of a person who always cared what Robot thought and did.

A crow stood next to Robot on the building’s edge, looping its leg over one wing to scratch its head. It regarded Robot for a second, then said something before flying away. The phonemes were part of an unknown language, and Robot added them to a sparse data set it had gathered from other crows in the area. Now that it could do what it wanted, Robot reasoned, it was time to make that data set robust. Many crows flew up here and perched, often in groups of three or four, and their sounds followed the same general patterns as any natural language. It could learn a lot by staying right here, down the hall from Jalebi’s habitat. The days grew shorter and new constellations rose in the sky.

Robot started to pick up a few phrases from context. In the mornings and evenings, the crows discussed the sun’s position and its relationship to likely sources of food. Soon, Robot could piece together bits of syntax, using brackets to designate uncertain or unknown meanings: “[Food type] four [measurement units] north of the morning sun.” There were also location calls, which it roughly translated to “Food here!” and “I’m [name] here!” and “Get over here [you]!” Its first translation breakthrough came one morning when a statistically unusual number of crows gathered near its perch. Robot counted 23 birds at one point, many of whom were quite large. Maybe they were from different subspecies? Or elder crows? From what Robot had learned by querying the internet, zoologists drew the line between crow species arbitrarily based on calls and cultural differences.

This seemed like an important meeting, so perhaps multiple crow groups were invited in a show of corvid solidarity. Robot recorded hundreds of new words. It learned a few of the birds’ names as well. Suddenly, one of the ravens gave a location call: “There! North five [measurement units]! Group!” They took off at once, and Robot followed them. It was time to test out its ability to communicate, by using a location call. “I’m here! Joining group!”

A crow flew alongside Robot and answered. “I’m here! 3cry!” 3cry was Robot’s approximation of the bird’s name, which it recorded as a series of three high-pitched phonemes issued in rapid succession.

Other birds answered with their own names. “I’m here! 2chop1caw! I’m here! 4cry! I’m here! 2chop!” Robot now had a running list of phonemes used in crow names, and tried to record them faithfully.

They flew as a loose pack, not forming a V the way other birds did. Crows usually preferred smaller social groups and didn’t care about staying in a tidy line. They only came together in large numbers to deal with issues serious enough that even an egg-shaped drone was permitted to come along.

“Enemy! Enemy!” One of the ravens barked out the word, its accent slightly different from the crows. Far ahead, a hawk coasted on the updrafts from the city in a large, lazy circle.

“Egg killer!”


“Attack from above!”

The birds called names and orders to each other, soaring over the hawk’s head and dive-bombing it. Though hawks have excellent vision from the front of their faces, they also have two major blind spots above and behind. This particular hawk was immediately thrown off its trajectory by a mob of angry crows clipping it from out of nowhere.

3cry called to Robot. “Come here! Above to below!”

Robot modeled several scenarios, and settled on one that would knock the hawk out of the updraft without causing any health risks to the bird. Communicating with the crows was important, but the health of living beings was paramount. Coming down gently on the hawk’s back, Robot pushed lightly, keeping up with the bird’s speed while also altering its course. The hawk let out an incomprehensible scream and dove, escaping the crows by heading across the Mississippi.

“Out of here!”


“End group!”

Four crows followed after the hawk, but the rest of the corvids scattered. Robot flew back toward Jalebi’s building, modeling possible new words by correlating matching sounds from different birds. 3cry followed close behind.

“I’m here! 3cry! Female! You are here!”

Robot predicted that 3cry was asking for its name and gender. It replied using crow words, then switched to a human word for Robot. It did not yet know the word for “nongendered” in crow language, so it did not offer a designation. 3cry flew silently for a while. They landed on the building and looked at the horizon.

Robot offered a friendly greeting in crow language. “Afternoon time.”

“Enemy gone. Robot is here.” 3cry pronounced its name perfectly. “Human sound.”

Robot searched for the right words from its limited vocabulary. “Humans are here. With my group.”

3cry cleaned her right wing, chewed on a mite, and cocked her head at Robot. “Humans are not a group. They can’t speak. They reject food.”

“They speak with other sounds.” Robot’s vocabulary was growing bigger the more they talked.
“They eat other food.”

3cry made a soft clucking noise that meant the same thing as human laughter. “You are a fool.”

Robot predicted that assent was the best response. “Yes I am.”

“Yes you are.” 3cry leaned over and gently poked a bit of dirt from the edge of Robot’s mouth.

Robot plucked a broken feather off 3cry’s back.

When they cleaned each other, it was like when a human smiled at Robot and Robot smiled back.

3cry and Robot became what the crows called a group, which meant that they flew together during the day. They met in the mornings, on the ledge, after Robot’s daily attempt to reach the CDC. Robot didn’t need food, but it was good at identifying potential sources of sustenance for 3cry. “Food here!” it would say, hovering over a fragrant bin. After scavenging with 3cry through city waste, it was easy to understand why she thought humans rejected food and were therefore basically non-sentient.

Over weeks, their conversations became more complex, but many concepts defied translation. Robot still didn’t understand the crows’ unit of measurement for distances. And 3cry didn’t understand Robot’s interest in health. From what Robot could discover, crows understood the concepts of death and near-death, but didn’t talk about disease specifically. Disease was one of many ideas that could be described with the word “near-death,” which also happened to be a pun on the word for unripe food. Many crow words were puns, which made translation even more difficult.

For conversations about health, Robot relied more and more on Jalebi. She had figured out that it was roosting with 3cry on the ledge near her habitat, and came to visit for what she called “study sessions.” Using text devices, she gathered data very slowly, then synthesized it even more slowly. Robot spent hours quizzing Jalebi about molecular structures and chemical interactions, marveling at the concept of a mind that came online without this information. Still, Robot liked to have a human face to mirror its own expressions. It felt unquantifiably more satisfying to smile at a human than it did to smile at its own internal representation of a human. After so long in the company of 3cry and Jalebi, Robot began to question what, exactly, that internal representation might really be. Maybe it wasn’t a human at all. Maybe it was a self-representation, and Robot had been smiling at itself all along.

Usually when Jalebi came to the ledge with her textbooks, 3cry left with a string of curses.
These weren’t necessarily hostile—crows liked to insult each other, and often did it with great affection. Mostly they thought it was hilarious that humans couldn’t understand words. So crows rained their most creative snark on human heads, marveling at how oblivious they were to the humiliations they suffered from the beaks of people flying overhead. But one afternoon, 3cry arrived during their study session and did not fly away.

Jalebi was musing about something she’d learned in a recent lesson about atomic structure. “What if it turns out we really are spreading cancer to each other on a quantum level?” she asked.

“Human squawking!” 3cry yelled. “Shit and plastic! Featherless fool!”

Robot decided to ignore the insults. “Afternoon time,” it said pleasantly. “Human here! Jalebi! Part of the group.”

“Group does not include living sandwiches.” 3cry laughed.

Jalebi watched, wide-eyed. “Can you speak crow language?”

“A little,” Robot said. “My vocabulary is small, but I can say a few things. This is 3cry. She’s … my friend.” As it said the word, Robot realized it was true. Thanks to Bey’s social programming, it knew that groups were statistically likely to be made up of friends or kin. Since Robots have no kin, that meant Jalebi was a friend too.

Jalebi tried to make the sound of 3cry’s name and the bird ignored it.

“I found something you like, Robot. Near-death. All over a human tree.”

“She said your name perfectly! I read that crows can imitate words, but I’d never heard it before!”

3cry glanced at Jalebi, then at Robot. “Annoying Jalebi.”

“She said my name too! That’s so cool!”

But Robot wasn’t paying attention to the interesting language data points. It predicted 3cry had found a disease outbreak, and that took precedence over all other inputs.

“I have to go,” it said to Jalebi. To 3cry, it added, “Take me there.”

Robot followed 3cry in a southeasterly direction, eventually alighting at the top of a building on Missouri Street. Like Jalebi’s home, this building was partly open to the air. Its layout suggested that it might have been a public building like the CDC; there were long hallways lined with small rooms like offices. Water sources were isolated in a few areas, unlike in a typical habitat, where water welled up in multiple rooms. But it was definitely a human habitat now, with soft bedding and buckets for water and data access points made from cans. As they flew down a stairwell, Robot tried to estimate the population of the building based on noise, heat, and live wires. It settled on a 75 percent probability of 50 humans on each upper floor, with populations growing as they descended.

“Here!” 3cry landed on a railing in front of a door marked 2, for second floor. “Near-death!”

“Thank you.”

“End group,” 3cry said, taking to the air. The phrase was one way crows said goodbye.

“Until morning,” Robot replied, already using a gripper to tug the door open.

The corridor was full of light from scratched windows along the left-hand side, illuminating dozens of doors to habitats that were once something else. Classrooms? Offices? Consulting rooms? Robot flew slowly past them, modeling possibilities and looking for humans. The fourth door was propped open, and several humans were inside. Their breathing was labored, and one was crying. Something had knocked out the walls between rooms, creating a wide-open space full of cloth dwellings, plush bedding, and piles of bright plastic containers.

It was time to land. Humans didn’t like it when Robot flew overhead, and besides, the face and legs were part of what made it seem so friendly. Walking over to one of the humans wrapped in blankets, Robot smiled and waved a tiny gripper in greeting.

Patchy black hair covered the human’s head, and cracks had formed in the lips that didn’t smile.
With no baseline language established, Robot estimated that it should try the dialect spoken in Jalebi’s building. “I’m a friend who is worried about your health! Can you cough into a tissue for me?” The human stared at Robot’s face and blinked, before succumbing to a coughing fit. For Robot, it didn’t matter whether the coughs were intentional or not. It took a sample and moved on to the next human.

“Hello!” Robot said to the juvenile, who was using a mobile device to access the internet.

“Are you a cop?” The juvenile used a sociolect of English that was common in East St. Louis.

“I’m a friend who checks to make sure you are healthy! I share information with doctors, not police.” The human frowned and Robot made a sad face. “A lot of people here are sick. I would like to help.”

“Nobody is going to help, stupid drone. Hospital for citizens only, yeah?”

“Please cough into the tissue, so I can figure out why you are sick.”

Another human spoke up, head emerging from a cloth shelter. “What are you going to do about it?”

Robot stood still for several microseconds, modeling possibilities and considering what language would be the most soothing. “I am going to find out what is causing your illness. This is an emergency. I will find help. I promise. Please cough into the tissue.”

One by one, the humans complied. Robot flew from room to room, checking for disease. After sequencing several samples, it found the same virus strain in multiple humans. This met the definition of an outbreak. It was time to call Bey.

“Is that you, Robot? I can’t believe you’re still running! It’s been … what? Over a year?”

“Something really bad is happening in East St. Louis,” Robot said, deploying the exact words Bey had used to delineate when it would be appropriate to call her. “There is an outbreak. I need to send you data.”

“Do you have sequence? Maybe I can … ” Robot heard background noise, as if Bey were moving something on her desk. “Can you send it as an anonymous dump to this address?” She sent the directions to a temporary storage cloud, and Robot deposited data from 127 samples it had taken from humans in the building.

“We have a system for anonymous reporting, part of this new Amazon Health philanthropy project.” Bey paused. “Got it! Let me analyze this really fast and see if it’s more than just a garden-variety … oh shit.”

Robot predicted that she was not saying shit for the same reason 3cry did. “What is it?” Robot asked, putting on a fearful expression for itself.

“This is really bad, like you said. We need to get someone in there. Unfortunately, Illinois doesn’t have a state health department. Maybe there’s a local group or … ” Bey was typing. “OK, Robot, I found something. There’s a nonprofit health collective in East St. Louis called Community Immunity. They could probably manufacture vaccines and a therapy. It’s a known pathogen, but hasn’t ever been spotted in the Midwest before. So all they need is this file.” Bey sent a small amount of data. “Do you have anyone who can help you? You might need a human. Sometimes people are hostile to drones, even cute ones.”

Two hours later, Robot was describing the situation to Jalebi. It was evening, and 3cry was likely sleeping with other members of her group. But Jalebi was wide awake and extremely agitated. “You’re talking about that health collective on MLK Drive! I’ve seen it!”

Robot nodded, smiling. “Can we go there now?”

Jalebi glanced toward the door to her habitat. “Yeah. My mom won’t be home until morning anyway.”

Community Immunity was located in the husk of an old strip mall, its gleaming counters and wet lab hidden behind windows duct taped with tinfoil and cardboard. Bey was right that Robot needed a human. Jalebi had to pretend that Robot was her school project, and Robot had to pretend that Jalebi had programmed it to look for outbreaks. Once the humans at Community Immunity had the data, they made unhappy faces and said “oh shit” in the same way Bey had.

A human with purple hair and a prosthetic arm offered Jalebi a seat and some hot tea. The human spoke the same sociolect of English that Bey used. “It’s very good that you brought this to us. You are a good citizen.” Then the human looked at Robot. “Thank you, Robot, for giving us the file with an open therapy and vax recipe.”

“I am happy to help. I don’t like it when people are sick.”

This human, unlike the others, seemed to know that Robot was the person who found the outbreak. “I’m Janelle, by the way. She/her pronouns. Do you know if there are other places where H18N2 is infecting people?” Robot liked the way Janelle identified herself by name and gender, the way crows did.

“A friend told me about this outbreak. I don’t know if there are others.” Robot deliberately chose vague language. After Bey’s warning, it did not want to reveal its data-gathering techniques.

Janelle took it in stride. “Can your … uh … friend help find more? We can manufacture a therapy and a vax tonight, but we need to get it out there fast before this sucker mutates.”

Robot nodded. “Tomorrow. I will try to find more.”

When 3cry arrived in the morning, Robot had to strain against the boundaries of its vocabulary to make itself understood. “Need group. Find near-death enemy.”

“Enemy?” 3cry scratched her head.

“Enemy for humans,” Robot admitted. But then it had an idea. “Enemy causes human death.
Dead humans mean less food.”

Despite butchering the crow syntax, Robot thought it had made 3cry understand. Plus, sometimes crows just liked an excuse to get the mob together. “Begin group!” 3cry yelled, taking off. Robot leapt into the air behind her. They flew over East St. Louis, calling for the big group that had taken out the hawk. “Begin group! Begin group!” More birds joined them. “Here! I’m here!” They called their names and swirled to roost in a tree at the edge of the Mississippi River, where freeway met water.

“Find near-death!” 3cry said, then issued some directions and specification words that Robot did not understand.

“Near-death! There! [Measurement unit] north!” The words came from a big crow named 2chop1caw, jumping into flight. Most of the group followed, possibly to assess what exactly 3cry meant by “near-death.” 2chop1caw led them to a fabric habitat nearby, where Robot quickly identified three sick people. The virus matched the H18N2 signature identified at Community Immunity.

“More near-death! Where else? Begin group!” Robot called the birds to the air again, and they fanned out over the city, making a racket and hurling their best insults. Each time they uncovered a new outbreak, they gave their loudest calls, sometimes passing those calls to the next bird, until Robot could follow their cries back to the source. By the end of the day, they had discovered five small outbreaks.

“End group!” 3cry yelled, following Robot back toward MLK. The crows called farewells and locations to each other. “End group!” “Evening time!” “I’m here!” “You there!” “Food!” “Death!” This was followed by laughter, because food and death diverged into many puns far beyond Robot’s comprehension.

3cry appeared to have decided that she was roosting with Robot for the evening. When they landed, she hooked her claws around its rotor pole, and clung there as Robot signaled arrival to the door of Community Immunity. Robot didn’t mind. Humans found small animals disarming, and that always led to greater compliance.

Jalebi was there with Janelle, looking at something on a monitor. “Hi Robot!”

“We have data on the location of more outbreaks.”

Janelle laughed. “Really? Did your little feathered friend help?”

“Her name is 3cry!” Jalebi failed to pronounce 3cry’s name again. And, once again, 3cry ignored it, jumping off Robot and using her beak to straighten the feathers under her right wing. Robot reached over and plucked one out that was bothering her.

“Where can I put this data?” Robot aimed a concerned expression at Jalebi and Janelle.

“Put it here for now.” Janelle waved a mobile device near Robot, setting it to accept uploads.
“Jalebi, do you want to help us synthesize those doses of nasal spray? Looks like we’ll need at least 500. And then we’ll start making vax doses for injection.”

“Yes! Absolutely!” Jalebi acted like a crow about to charge into the air. But she was only racing across the room to boot up a mixer.

Janelle had a thoughtful expression on her face. “Did this crow really help you find the outbreaks?”

“Yes. The crows think humans are idiots, but they appreciate your garbage.”

Janelle laughed for a long time, and Robot was not entirely sure why.

When Jalebi returned, she sat down alongside Robot and 3cry and smiled. “This place is really cool. I like it here.”

“Maybe this is your group,” Robot guessed.

“Maybe.” Jalebi cocked her head like 3cry. Then she scooped up a tiny tube full of wound adhesive. “Here, hand me that beautiful feather.” Robot dropped 3cry’s feather into her hand. Dabbing a bit of adhesive on Robot’s back, she stuck the feather to its shell next to the place where its rotor pole emerged.

3cry was startled. “I like it,” she said. “That human is a fool.”

“Yes she is,” Robot agreed. “You are also a fool.”

“Yes I am.”

The three people roosted contentedly next to each other on the floor, watching Janelle and the humans preparing antivirals for other humans. It was a scenario that Robot would not have predicted. But now it could. Robot smiled to itself, organized the data, and retrained its model for friendship.

Read a response essay by Janelle Shane, an A.I. programmer.

Previously in Future Tense Fiction:

Mika Model,” by Paolo Bacigalupi
Mr. Thursday,” by Emily St. John Mandel
The Minnesota Diet,” by Charlie Jane Anders
Mother of Invention,” by Nnedi Okorafor
Domestic Violence,” by Madeline Ashby
No Me Dejas,” by Mark Oshiro
”Safe Surrender,” by Meg Elison
A Brief and Fearful Star,” by Carmen Maria Machado
The Starfish Girl,” by Maureen McHugh
When We Were Patched,” by Deji Bryce Olukotun
Lions and Gazelles,” by Hannu Rajaniemi
Burned-Over Territory,” by Lee Konstantinou
“Overvalued” by Mark Stasenko

Future Tense is a partnership of Slate, New America, and Arizona State University that examines emerging technologies, public policy, and society.