Computers Math

How could the fictitious Marvel movie character Ant-Man produce high energy out of his small body? The secret lies in the “transistors” on his suit that amplify weak signals for processing. Transistors that amplify electrical signals in the conventional way lose heat energy and limit the speed of signal transfer, which degrades performance. What if it were possible to overcome such limitation and make a high-performance suit that is light and small but without loss of heat energy?
A POSTECH team of Professor Kyoung-Duck Park and Yeonjeong Koo from the Department of Physics and a team from ITMO University in Russia led by Professor Vasily Kravtsov jointly developed a “nano-excitonic transistor” using intralayer and interlayer excitons in heterostructure-based semiconductors, which addresses the limitationsof existing transistors.
“Excitons” are responsible for light emission of semiconductor materials and are key to developing a next-generation light-emitting element with less heat generation and a light source for quantum information technology due to the free conversion between light and material in their electrically neutral states. There are two types of excitons in a semiconductor heterobilayer, which is a stack of two different semiconductor monolayers: the intralayer excitons with horizontal direction and the interlayer excitons with vertical direction.
Optical signals emitted by the two excitons have different lights, durations, and coherence times. This means that selective control of the two optical signals could enable the development of a two-bit exciton transistor. However, it was challenging to control intra- and interlayer excitons in nano-scale spaces due to the non-homogeneity of semiconductor heterostructures and low luminous efficiency of interlayer excitons in addition to the diffraction limit of light.
The team in its previous research had proposed technology for controlling excitons in nano-level spaces by pressing semiconductor materials with a nano-scale tip. This time, for the first time ever, the researchers were able to remotely control the density and luminance efficiency of excitons based on polarized light on the tip without directly touching the excitons. The most significant advantage of this method, which combines a photonic nanocavity and a spatial light modulator, is that it can reversibly control excitons, minimizing physical damage to the semiconductor material. Also, the nano-excitonic transistor that utilizes “light” can help process massive amounts of data at the speed of light while minimizing heat energy loss.
Artificial intelligence (AI) has made inroads into our lives more quickly than we ever expected, and it requires huge volumes of data for learning in order to provide good answers that are actually helpful for users. The ever-increasing volume of information should be collected and processed as more and more fields utilize AI. This research is expected to propose a new data processing strategy befitting an era of data explosion. Yeonjeong Koo, one of the co-first authors of the research paper, said, “The nano-excitonic transistor is expected to play an integral role in realizing an optical computer, which will help process the huge amounts of data driven by AI technology.
The research, recently published in international journal ACS Nano, was supported by the Samsung Science and Technology Foundation and National Research Foundation of Korea. More

The flow of matter, from macroscopic water currents to the microscopic flow of electric charge, underpins much of the infrastructure of modern times. In the search for breakthroughs in energy efficiency, data storage capacity, and processing speed, scientists search for ways in which to control the flow of quantum aspects of matter such as the “spin” of an electron – its magnetic moment – or its “valley state”, a novel quantum aspect of matter found in many two dimensional materials. A team of researchers at the Max Born Institute in Berlin have recently discovered a route to induce and control the flow of spin and valley currents at ultrafast times with specially designed laser pulses, offering a new perspective on the ongoing search for the next generation of information technologies.
Ultrafast laser control over the fundamental quantum degrees of freedom of matter represents the outstanding foundational challenge to be met in establishing future information technologies beyond the semi-conductor electronics that defines our present time. Two of the most promising quantum degrees of freedom in this respect are the spin of the electron and the “valley index,” the latter an emergent degree of freedom of two dimensional materials related to the quasiparticle momentum. Both spintronics and valleytronics offer many potential advantages over classical electronics in terms of data manipulation velocity and energy efficiency. However, while spin excitations suffer from a dynamical loss of character arising from the spin-orbit induced spin precession, the valley wavefunction represents a “data bit” whose stability is threatened only by intervalley scattering, a feature controllable be sample quality. Valleytronics thus presents a potentially robust platform for going beyond classical electronics.
At the heart of any future valleytronics or spintronics technologies will, in addition to quantum excitations encoding data bits, reside the control and creation of valley- and spin-currents. However, while sustained attention has been paid to the task of tailoring lightforms on ultrafast time scales to selectively excite valley quasiparticles, the precise creation and control of valley-currents and spin-currents — vital for any future valleytronics technology — has remained beyond the realm of ultrafast light control. In a study recently published in Science Advances, a team of researchers from the Max Born Institute in Berlin have shown how a hybrid laser pulse combining two polarization types allows complete control over ultrafast laser-light-induced currents.
Control over the charge state by circularly polarized light is now well established, the famous “spin-valley locking” of the transition metal dichalcogenides that has its origin in the valley selective response to circularly polarized light. This can be viewed as arising from a selection rule involving the magnetic quantum numbers of the d-orbitals that comprise the gap edge states. While circularly polarized light excites valley charge it does not, however, create a valley current. This situation arises as for each quasi-momentum in the valley
Control over the charge state by circularly polarized light is now well established, the famous “spin-valley locking” of the transition metal dichalcogenides that has its origin in the valley selective response to circularly polarized light. This can be viewed as arising from a selection rule involving the magnetic quantum numbers of the d-orbitals that comprise the gap edge states. While circularly polarized light excites valley charge it does not, however, create a valley current. This situation arises as for each quasi-momentum in the valley kvalley that is excited a corresponding -kvalley also is excited: the Bloch velocities thus cancel and there is no net valley current.
Full control over light induced valley currents, their magnitude and direction, thus requires going beyond the spin-valley locking paradigm of circularly polarized light. Creation of a valley excited state that does result in a net valley and spin current must therefore involving breaking the local kvalley , -kvalley degeneracy. As the laser vector potential couples directly to crystal quasi-momentum, k – >k — A (t)/c, the most effective way in which this can be done is through a linearly polarized single cycle pulse with duration comparable to that of the circularly polarized pulse: such a pulse will evidently be in the “THz window” of 1 THz to 50 THz. The hencomb lightform generates a substantial residual (i.e. persisting after the laser pulse) current. This results from a non-cancellation of the Bloch velocities of excited quasi-momentum, as the distribution of excited charge is now shifted off the high symmetry K point by exactly the polarization vector of the THz pulse,. More

In recent years, robots have become incredibly sophisticated machines capable of performing or assisting humans in all tasks. The days of robots functioning behind a security barrier are long gone, and today we may anticipate robots working alongside people in close contact. While working alongside robots may be very practical in some situations, they should be designed to be safe and pleasant for humans to interact with. For instance, in human-robot interactions (HRIs), robots should be able to react correctly to potential collisions with humans and also respond safely and predictably to intentional physical contact.
One of the best approaches to improve HRIs is to grant robots the ability to sense their environment in multiple ways, such as by touch, sound, and sight. Of these three, tactile sensation is particularly important for robots that are likely to come into physical contact with humans during operation. Although small-scale tactile sensors have seen tremendous progress over the past decade, the development of large-scale tactile sensors has been plagued with challenges. Moreover, most researchers have focused on systems that respond to physical touch and ignore touchless stimuli, such as when an object is in close proximity. To address these issues, a research team led by Associate Professor Van Anh Ho from Japan Advanced Institute of Science and Technology (JAIST) recently developed ProTac — an innovative soft robotic link with tactile and proximity sensing capabilities. As explained in their paper, presented at the IEEE-RAS International Conference on Soft Robotics (ROBOSOFT), the team not only engineered ProTac itself but also pioneered a new simulation and learning framework to effectively prepare the robotic link for use.
But what does a robotic link look like, and what is ProTac good for? In general, robotic links are rigid structural components of a robot that connect two or more joints. For example, robotic links can be seen as various ‘segments’ in a robotic limb. In this study, ProTac is designed as a soft, cylindrical segment for a robotic arm. What makes it remarkable is how the researchers incorporated the tactile and proximity sensing capabilities in a very convenient and space-efficient way.
ProTac has an outer ‘soft magic skin’ that can be slightly deformed by touch without damage. The inside of the skin is patterned with arrays of reflective markers, and fisheye cameras are installed at both ends of the robotic link looking towards these markers. The idea is that, upon physical contact and deformation of the skin, changes in the relative positions of the markers are captured by the cameras and processed to calculate the precise location and intensity of the contact. On top of this, the outer skin is of a functional polymer that can be made entirely transparent by applying an external voltage. It allows the fisheye cameras to image the immediate surroundings of ProTac, providing footage for proximity calculations.
To more easily train ProTac to make proximity and tactile measurements, the team also developed SimTacLS, an open-source simulation and learning framework based on the SOFA and Gazebo physics engines (see the paper here). This machine learning framework is trained with simulated and experimental data considering the physics of soft contact and the realistic rendering of sensor images. “SimTacLS enabled us to effectively implement tactile perception in robotic links without the high costs of complex experimental setups,” remarks Prof. Ho, “Furthermore, with this framework, users can readily validate sensor designs and learning-based sensing performance before proceeding to actual fabrication and implementation.”
Overall, this work will help pave the way to a world where humans can harmoniously coexist and work alongside robots. Excited by the team’s contribution to this dream, Prof. Ho comments: “We expect the proposed sensing device and framework to bring in ultimate solutions for the design of robots with softness, whole-body and multimodal sensing, and safety control strategies.” It is worth noting that proposed techniques can be extended to other types of robotic systems beyond the robotic manipulator demonstrated in the study, such as mobile and flying robots. Moreover, ProTac or similar robotic links could be used to enable robotic manipulation in cluttered environments or when operating in close vicinity with humans. More

The field of integrated photonics has taken off in recent years. These microchips utilise light particles (photons) in their circuitry as opposed to the electronic circuits that, in many ways, form the backbone of our modern age. Offering improved performance, reliability, energy efficiency, and novel functionalities, integrated photonics has immense potential and is fast becoming a part of the infrastructure in data centres and telecom systems, while also being a promising contender for a wide range of sensors and integrated quantum technologies.
Significant improvements in nanoscale fabrication have made it possible to build photonic circuits with minimal defects, but defects can never be entirely avoided, and losses due to disorder remains a limiting factor in today’s technology. Minimising these losses could, for example, reduce the energy consumption in communication systems and further improve the sensitivity of sensor technology. And since photonic quantum technologies rely on encoding information in fragile quantum states, minimising losses is essential to scale quantum photonics to real applications. So the search is on for new ways to reduce the backscattering, or even prevent it entirely.
A one-way street for photons is impossible today
One suggestion for minimising the loss of photons in an integrated photonic system is to guide the light through the circuit using topological interfaces that prevent backscattering by design.
“It would be very nice if it were possible to reduce losses in these systems. But fundamentally, creating such a one-way street for photons is a tough thing to do. In fact, as of right now, it is impossible; to do this in the optical domain would require developing new materials that do not exist today,” says Associate Professor Søren Stobbe, Group Leader at DTU Electro.
Circuitry built from topological insulators would, in theory, force photons to keep moving forward, never backward. The backwards channel would simply not exist. While such effects are well-known in niche electronics and have been demonstrated with microwaves, they have yet to be shown in the optical domain.

But full topological protection is impossible in silicon and all other low-loss photonic materials, because they are subject to time-reversal symmetry. This means that whenever a waveguide allows transmitting light in one direction, the backwards path is also possible. This means that there is no one-way street for photons in conventional materials, but researchers have hypothesized that a two-way street would already be good enough to prevent backscattering.
“There has been a lot of work trying to realise topological waveguides in platforms relevant for integrated photonics. One of the most interesting platforms is silicon photonics, which uses the same materials and technology that make up today’s ubiquity of computer chips to build photonic systems, and even if disorder cannot be entirely eliminated, perhaps backscattering can,” says Søren Stobbe.
New experimental results from DTU recently published in Nature Photonics strongly suggest that with the materials available today, this likely will not happen.
State-of-the-art waveguides offer no protection
Although several previous studies have found that it may be possible to prevent backscattering based on various indirect observations, rigorous measurements of the losses and the backscattering in topological waveguides were so far missing. The central experiments conducted at DTU were performed on a highly well-characterised state-of-the-art type of silicon waveguide, showing that even in the best waveguides available, the topological waveguides show no protection against backscattering.

“We fabricated the best waveguide obtainable with current technology — reporting the smallest losses ever seen and reaching minute levels of structural disorder — but we never saw topological protection against backscattering. If the two-way topological insulators protect against backscattering, they would only be effective at disorder levels below what is possible today,” says PhD-student Christian Anker Rosiek.
He conducted most of the fabrication, experiments and data analysis along with postdoc Guillermo Arregui, both at DTU Electro.
“Measuring the losses alone is crucial, but not enough, because losses can also come from radiation out of the waveguide. We can see from our experiments that the photons get caught in little randomly located cavities in the waveguide as if many of tiny mirrors had been randomly placed in the light’s path. Here, the light is reflected back and forth, scattering very strongly on those defects. It shows that the backscattering strength is high, even in a state-of-the-art system, proving that backscattering is the limiting factor,” says Guillermo Arregui.
Waveguide-material should break time-reversal symmetry
The study concludes that, for a waveguide to offer protection against backscattering, you would need the topological insulator to be constructed from materials that break time-reversal symmetry without absorbing light. Such materials do not exist today.
“We are not ruling out that protection from backscattering can work, and absence of evidence must not be confused with evidence of absence. There is plenty of exciting research to be explored within topological physics, but moving forward, I believe researchers should take great care in measuring losses when presenting new topological waveguides. That way, we will get a clearer picture of the true potential of these structures. Suppose someone does indeed develop new, exotic materials that allow only propagation in one direction, our study has established the tests needed to claim real protection against backscattering.,” says Christian Anker Rosiek. More

Researchers have designed a low-cost, energy-efficient robotic hand that can grasp a range of objects — and not drop them — using just the movement of its wrist and the feeling in its ‘skin’.
Grasping objects of different sizes, shapes and textures is a problem that is easy for a human, but challenging for a robot. Researchers from the University of Cambridge designed a soft, 3D printed robotic hand that cannot independently move its fingers but can still carry out a range of complex movements.
The robot hand was trained to grasp different objects and was able to predict whether it would drop them by using the information provided from sensors placed on its ‘skin’.
This type of passive movement makes the robot far easier to control and far more energy-efficient than robots with fully motorised fingers. The researchers say their adaptable design could be used in the development of low-cost robotics that are capable of more natural movement and can learn to grasp a wide range of objects. The results are reported in the journal Advanced Intelligent Systems.
In the natural world, movement results from the interplay between the brain and the body: this enables people and animals to move in complex ways without expending unnecessary amounts of energy. Over the past several years, soft components have begun to be integrated into robotics design thanks to advances in 3D printing techniques, which have allowed researchers to add complexity to simple, energy-efficient systems.
The human hand is highly complex, and recreating all of its dexterity and adaptability in a robot is a massive research challenge. Most of today’s advanced robots are not capable of manipulation tasks that small children can perform with ease. For example, humans instinctively know how much force to use when picking up an egg, but for a robot this is a challenge: too much force, and the egg could shatter; too little, and the robot could drop it. In addition, a fully actuated robot hand, with motors for each joint in each finger, requires a significant amount of energy.

In Professor Fumiya Iida’s Bio-Inspired Robotics Laboratory in Cambridge’s Department of Engineering, researchers have been developing potential solutions to both problems: a robot hand than can grasp a variety of objects with the correct amount of pressure while using a minimal amount of energy.
“In earlier experiments, our lab has shown that it’s possible to get a significant range of motion in a robot hand just by moving the wrist,” said co-author Dr Thomas George-Thuruthel, who is now based at University College London (UCL) East. “We wanted to see whether a robot hand based on passive movement could not only grasp objects, but would be able to predict whether it was going to drop the objects or not, and adapt accordingly.”
The researchers used a 3D-printed anthropomorphic hand implanted with tactile sensors, so that the hand could sense what it was touching. The hand was only capable of passive, wrist-based movement.
The team carried out more than 1200 tests with the robot hand, observing its ability to grasp small objects without dropping them. The robot was initially trained using small 3D printed plastic balls, and grasped them using a pre-defined action obtained through human demonstrations.
“This kind of hand has a bit of springiness to it: it can pick things up by itself without any actuation of the fingers,” said first author Dr Kieran Gilday, who is now based at EPFL in Lausanne, Switzerland. “The tactile sensors give the robot a sense of how well the grip is going, so it knows when it’s starting to slip. This helps it to predict when things will fail.”
The robot used trial and error to learn what kind of grip would be successful. After finishing the training with the balls, it then attempted to grasp different objects including a peach, a computer mouse and a roll of bubble wrap. In these tests, the hand was able to successfully grasp 11 of 14 objects.

“The sensors, which are sort of like the robot’s skin, measure the pressure being applied to the object,” said George-Thuruthel. “We can’t say exactly what information the robot is getting, but it can theoretically estimate where the object has been grasped and with how much force.”
“The robot learns that a combination of a particular motion and a particular set of sensor data will lead to failure, which makes it a customisable solution,” said Gilday. “The hand is very simple, but it can pick up a lot of objects with the same strategy.”
“The big advantage of this design is the range of motion we can get without using any actuators,” said Iida. “We want to simplify the hand as much as possible. We can get lots of good information and a high degree of control without any actuators, so that when we do add them, we’ll get more complex behaviour in a more efficient package.”
A fully actuated robotic hand, in addition to the amount of energy it requires, is also a complex control problem. The passive design of the Cambridge-designed hand, using a small number of sensors, is easier to control, provides a wide range of motion, and streamlines the learning process.
In future, the system could be expanded in several ways, such as by adding computer vision capabilities, or teaching the robot to exploit its environment, which would enable it to grasp a wider range of objects.
This work was funded by UK Research and Innovation (UKRI), and Arm Ltd. Fumiya Iida is a Fellow of Corpus Christi College, Cambridge. More

Researchers have developed a new chip-sized microwave photonic filter to separate communication signals from noise and suppress unwanted interference across the full radio frequency spectrum. The device is expected to help next-generation wireless communication technologies efficiently convey data in an environment that is becoming crowded with signals from devices such as cell phones, self-driving vehicles, internet-connected appliances and smart city infrastructure.
“This new microwave filter chip has the potential to improve wireless communication, such as 6G, leading to faster internet connections, better overall communication experiences and lower costs and energy consumption for wireless communication systems,” said researcher Xingjun Wang from Peking University. “These advancements would directly and indirectly affect daily life, improving overall quality of life and enabling new experiences in various domains, such as mobility, smart homes and public spaces.”
In the Photonics Research journal co-published by Chinese Laser Press and Optica Publishing Group, the researchers describe how their new photonic filter overcomes the limitations of traditional electronic devices to achieve multiple functionalities on a chip-sized device with low power consumption. They also demonstrate the filter’s ability to operate across a broad radio frequency spectrum extending to over 30 GHz, showing its suitability for envisioned 6G technology.
“As the electro-optic bandwidth of optoelectronic devices continues to increase unstoppably, we believe that the integrated microwave photonics filter will certainly be one of the important solutions for future 6G wireless communications,” said Wang. “Only a well-designed integrated microwave photonics link can achieve low cost, low power consumption and superior filtering performance.”
Stopping interference
6G technology is being developed to improve upon currently-deployed 5G communications networks. To convey more data at a faster rate, 6G networks are expected to use millimeter wave and even terahertz frequency bands. As this will distribute signals over an extremely wide frequency spectrum with increased data rate, there is a high likelihood of interference between different communication channels.

To solve this problem, researchers have sought to develop a filter that can protect signal receivers from various types of interference across the full radio frequency spectrum. To be cost-effective and practical for widespread deployment, it is important for this filter to be small, consume little power, achieve multiple filtering functions and be able to be integrated on a chip. However, previous demonstrations have been limited by their few functions, large size, limited bandwidth or requirements associated with electrical components.
For the new filter, researchers created a simplified photonic architecture with four main parts. First, a phase modulator serves as the input of the radio frequency signal, which modulates the electrical signal onto the optical domain. Next, a double-ring acts as a switch to shape the modulation format. An adjustable microring is the core unit for processing the signal. Finally, a photodetector serves as the output of the radio frequency signal and recovers the radio frequency signal from the optical signal.
“The greatest innovation here is breaking the barriers between devices and achieving mutual collaboration between them,” said Wang. “The collaborative operation of the double-ring and microring enables the realization of the intensity-consistent single-stage-adjustable cascaded-microring (ICSSA-CM) architecture. Owing to the high reconfigurability of the proposed ICSSA-CM, no extra radio frequency device is needed for the construction of various filtering functions, which simplifies the whole system composition.”
Demonstrating performance
To test the device, researchers used high-frequency probes to load a radio frequency signal into the chip and collected the recovered signal with a high-speed photodetector. They used an arbitrary waveform generator and directional antennas to simulate the generation of 2Gb/s high-speed wireless transmission signals and a high-speed oscilloscope to receive the processed signal. By comparing the results with and without the use of the filter, the researchers were able to demonstrate the filter’s performance.
Overall, the findings show that the simplified photonic architecture achieves comparable performance with lower loss and system complexity compared with previous programmable integrated microwave photonic filters composed of hundreds of repeating units. This makes it more robust, more energy-efficient and easier to manufacture than previous devices.
The researchers plan to further optimize the modulator and improve the overall filter architecture to achieve a high dynamic range and low noise while ensuring high integration at both the device and system levels. More

Captain of her high school tennis team and a four-year veteran of varsity tennis in college, Amanda Studnicki had been training for this moment for years.
All she had to do now was think small. Like ping pong small.
For weeks, Studnicki, a graduate student at the University of Florida, served and rallied against dozens of players on a table tennis court. Her opponents sported a science-fiction visage, a cap of electrodes streaming off their heads into a backpack as they played against either Studnicki or a ball-serving machine. That cyborg look was vital to Studnicki’s goal: to understand how our brains react to the intense demands of a high-speed sport like table tennis – and what difference a machine opponent makes.
Studnicki and her advisor, Daniel Ferris, discovered that the brains of table tennis players react very differently to human or machine opponents. Faced with the inscrutability of a ball machine, players’ brains scrambled themselves in anticipation of the next serve. While with the obvious cues that a human opponent was about to serve, their neurons hummed in unison, seemingly confident of their next move.
The findings have implications for sports training, suggesting that human opponents provide a realism that can’t be replaced with machine helpers. And as robots grow more common and sophisticated, understanding our brains’ response could help make our artificial companions more naturalistic.
“Robots are getting more ubiquitous. You have companies like Boston Dynamics that are building robots that can interact with humans and other companies that are building socially assistive robots that help the elderly,” said Ferris, a professor of biomedical engineering at UF. “Humans interacting with robots is going to be different than when they interact with other humans. Our long term goal is to try to understand how the brain reacts to these differences.”
Ferris’s lab has long studied the brain’s response to visual cues and motor tasks, like walking and running. He was looking to upgrade to studying complex, fast-paced action when Studnicki, with her tennis background, joined the research group. So the lab decided tennis was the perfect sport to address these questions with. But the oversized movements – especially high overhand serves – proved an obstacle to the burgeoning tech.

“So we literally scaled things down to table tennis and asked all the same questions we had for tennis before,” Ferris said. The researchers still had to compensate for the smaller movements of table tennis. So Ferris and Studnicki doubled the 120 electrodes in a typical brain-scanning cap, each bonus electrode providing a control for the rapid head movements during a table tennis match.
With all these electrodes scanning the brain activity of players, Studnicki and Ferris were able to tune into the brain region that turns sensory information into movement. This area is known as the parieto-occipital cortex.
“It takes all your senses – visual, vestibular, auditory – and it gives information on creating your motor plan. It’s been studied a lot for simple tasks, like reaching and grasping, but all of them are stationary,” Studnicki said. “We wanted to understand how it worked for complex movements like tracking a ball in space and intercepting it, and table tennis was perfect for this.”
The researchers analyzed dozens of hours of play against both Studnicki and the ball machine. When playing against another human, players’ neurons worked in unison, like they were all speaking the same language. In contrast, when players faced a ball-serving machine, the neurons in their brains were not aligned with one another. In the neuroscience world, this lack of alignment is known as desynchronization.
“If we have 100,000 people in a football stadium and they’re all cheering together, that’s like synchronization in the brain, which is a sign the brain is relaxed” Ferris said. “If we have those same 100,000 people but they’re all talking to their friends, they’re busy but they’re not in sync. In a lot of cases, that desynchronization is an indication that the brain is doing a lot of calculations as opposed to sitting and idling.”
The team suspects that the players’ brains were so active while waiting for robotic serves because the machine provides no cues of what they are going to do next. What’s clear is that our brains process these two experiences very differently, which suggests that training with a machine might not offer the same experience as playing against a real opponent.
“I still see a lot of value in practicing with a machine,” Studnicki said. “But I think machines are going to evolve in the next 10 or 20 years, and we could see more naturalistic behaviors for players to practice against.” More

Most kids know it’s wrong to yell or hit someone, even if they don’t always keep their hands to themselves. But what about if that someone’s name is Alexa?
A new study from Duke developmental psychologists asked kids just that, as well as how smart and sensitive they thought the smart speaker Alexa was compared to its floor-dwelling cousin Roomba, an autonomous vacuum.
Four- to eleven-year-olds judged Alexa to have more human-like thoughts and emotions than Roomba. But despite the perceived difference in intelligence, kids felt neither the Roomba nor the Alexa deserve to be yelled at or harmed. That feeling dwindled as kids advanced towards adolescence, however. The findings appear online April 10 in the journal Developmental Psychology.
The research was inspired in part by lead author Teresa Flanagan seeing how Hollywood depicts human-robot interactions in shows like HBO’s “Westworld.”
“In Westworld and the movie Ex Machina, we see how adults might interact with robots in these very cruel and horrible ways,” said Flanagan, a visiting scholar in the department of psychology & neuroscience at Duke. “But how would kids interact with them?”
To find out, Flanagan recruited 127 children aged four to eleven who were visiting a science museum with their families. The kids watched a 20-second clip of each technology, and then were asked a few questions about each device.

Working under the guidance of Tamar Kushnir, Ph.D., her graduate advisor and a Duke Institute for Brain Sciences faculty member, Flanagan analyzed the survey data and found some mostly reassuring results.
Overall, kids decided that both the Alexa and Roomba probably aren’t ticklish and wouldn’t feel pain if they got pinched, suggesting they can’t feel physical sensations like people do. However, they gave Alexa, but not the Roomba, high marks for mental and emotional capabilities, like being able to think or getting upset after someone is mean to it.
“Even without a body, young children think the Alexa has emotions and a mind,” Flanagan said. “And it’s not that they think every technology has emotions and minds — they don’t think the Roomba does — so it’s something special about the Alexa’s ability to communicate verbally.”
Regardless of the different perceived abilities of the two technologies, children across all ages agreed it was wrong to hit or yell at the machines.
“Kids don’t seem to think a Roomba has much mental abilities like thinking or feeling,” Flanagan said. “But kids still think we should treat it well. We shouldn’t hit or yell at it even if it can’t hear us yelling.”
The older kids got however, the more they reported it would be slightly more acceptable to attack technology.

“Four- and five-year-olds seem to think you don’t have the freedom to make a moral violation, like attacking someone,” Flanagan said. “But as they get older, they seem to think it’s not great, but you do have the freedom to do it.”
The study’s findings offer insights into the evolving relationship between children and technology and raise important questions about the ethical treatment of AI and machines in general, and as parents. Should adults, for example, model good behavior for their kids by thanking Siri or its more sophisticated counterpart ChatGPT for their help?
For now, Flanagan and Kushnir are trying to understand why children think it is wrong to assault home technology.
In their study, one 10-year-old said it was not okay to yell at the technology because, “the microphone sensors might break if you yell too loudly,” whereas another 10-year-old said it was not okay because “the robot will actually feel really sad.”
“It’s interesting with these technologies because there’s another aspect: it’s a piece of property,” Flanagan said. “Do kids think you shouldn’t hit these things because it’s morally wrong, or because it’s somebody’s property and it might break?”
This research was supported by the U.S. National Science Foundation (SL-1955280, BCS-1823658). More