When the Body Becomes a Communication Channel
On signals, understanding, and the responsibility that remains
A parent stands beside their baby’s bed. The baby is asleep, wearing a small sock around one foot. Inside the sock is a sensor. On the phone, figures for pulse rate and oxygen saturation appear alongside information about sleep. The parent looks at the baby, then at the screen.
Both provide information. But which has the final word?
This is an imagined scene, but the technology exists. Small devices monitor the body and transmit information. In an article in Science, published on 24 September 2026, I encounter another, more experimental connection: electronic devices send signals through the body’s own tissue, activating other devices inside it.
I pause over the word communication. What has happened when a signal arrives? Has anyone understood anything? And who must answer if the connection fails?
What is new
Ramy Ghanim and his colleagues have developed SWANS, the Smart Wireless Autonomous Networking System. It takes advantage of the fact that bodily tissue conducts electrical current through ions. Pulses with particular characteristics activate particular electronic circuits, allowing several devices to form part of the same network (Ghanim et al., 2026).
Ionic communication through tissue is not itself new. The article refers, among other examples, to Abilify MyCite, in which an ingested sensor transmits information to a patch worn on the body. SWANS contributes a combination of selective activation of multiple devices, very low power consumption while implants await signals, small dimensions, and a long communication range in tissue samples. Implants can be made small enough to be inserted using a syringe.
The researchers investigated the system in tissue samples and living rats. In the demonstration of nerve stimulation, strain sensors on the forelimbs were connected by wires to a hub on the animal’s back. The hub processed the sensor data and transmitted pulses wirelessly through the tissue to implants near the hindlimbs. These implants stimulated the sciatic nerves and produced movement.
The study also included two months of testing in living rats to investigate safety and communication over time. No humans were studied. The results demonstrate a promising technical connection, while further investigation remains necessary before any use in humans.
When a threshold becomes an intention
The article describes the strain sensors as detecting motion intent. I notice the choice of words.
The sensors register a physical change that produces an electrical response. When the signal crosses a threshold, the hub sends a pulse onwards. “Intent” is the term the researchers give this registration within the system.
A movement can serve as an indication of intended activity. But the study does not thereby give us access to an intention experienced by the animal. Between the sensor reading and the word intention lies an interpretation.
I recognise a related leap in my own hearing. When a word does not reach me clearly, I try to find meaning in the sounds I do hear. Context helps, as does my expectation of what the other person is likely to say. Sometimes I get it right. At other times, I have understood something that was never said.
Then I need the opportunity to ask: What did you mean?
For Hans-Georg Gadamer, we encounter the world with experiences and expectations that make understanding possible. Yet these must also be open to challenge by what we encounter. In conversation, the other person can correct my initial interpretation, and the matter we are discussing can reveal itself differently from what I expected (Gadamer, 1960/2004).
The sensor cannot answer when we ask what the word intention means here. We must address that question to the people interpreting its signal.
Listening and confirming
The authors also use the expression listening state. The implant “listens” while awaiting a signal, consuming very little power. The metaphor describes a technical property, but it brings with it the image of someone who is attentive and ready to hear.
The implant is receptive to particular electrical pulses. It responds within the conditions built into its circuit, without assessing the purpose of the action.
Claude Shannon’s theory of communication helps us understand the distinction. He investigated how messages could be transmitted through a channel, while setting questions of meaning aside from the technical task (Shannon, 1948).
In SWANS, the message can be very brief: a pulse that activates a particular device. The researchers emphasise that the system is suited to small amounts of data and relatively infrequent activation. It trades bandwidth for miniaturisation and longevity.
The article also describes how an implant can send back a signal confirming its activation. The system thus receives information that something has happened at the receiving end. This is an important check, but the confirmation does not establish whether the action was appropriate or what it meant for the person living with the system.
A person’s response can change our understanding of the matter itself. A confirmation signal tells us that an event has occurred within the system’s predefined framework.
When the body changes the channel
The body is not an unchanging wire. During the two months of testing, connective tissue formed around the implants, and electrical impedance at the receiving pads increased. The researchers maintained communication by increasing the voltage within limits they considered safe under the experimental conditions. Differences between the rats also required individual settings (Ghanim et al., 2026).
The connection therefore depends on a body that responds to the device and changes over time. A setting that works today may need adjustment later.
Who monitors this shift? Who decides when the signal strength should be increased and when the system should instead be examined or stopped?
The study describes how the researchers handled this during the experiments. If the technology is eventually used in treatment, responsibility for its continued monitoring must also be clearly assigned.
Another demonstration shows how an implant containing a temperature-sensitive resistor responds at temperatures above 40°C and sends a triggering signal to another implant. The researchers describe this as detecting a high fever. The demonstration does not document the treatment of a patient with fever, but it shows how a registered bodily condition can trigger an action without human assessment at that moment.
The assessment has moved into the design. Someone has selected the threshold, decided what it should trigger, and assessed which errors are acceptable. Automation embeds earlier decisions in an arrangement that subsequently acts.
Many hands
Dennis Thompson described “the problem of many hands”: when many people contribute to a decision or outcome, it becomes difficult to assign moral responsibility to particular individuals (Thompson, 1980). Helen Nissenbaum examined this problem in computer systems, showing how complexity, software errors, and blaming the computer can obstruct accountability (Nissenbaum, 1996).
A medical network involves many such hands. Some develop the sensor. Others write the software, select thresholds, assess safety, recommend the product, or support the user.
The person wearing the device does not necessarily know about these choices or have any influence over them. Using a system does not make someone responsible for everything it does.
Ethically, responsibility must correspond to knowledge, role, and capacity to act. The manufacturer must answer for the design and make its limitations understandable. Healthcare professionals must answer for clinical judgements within their roles. Users must receive instructions and opportunities to respond that match what they are expected to do.
The device forms part of the chain of causes, but it cannot itself answer morally for that chain. Shared responsibility therefore requires clear connections between those responsible. Otherwise, responsibility risks disappearing precisely where communication becomes most sophisticated.
When gaps acquire content
The question of reliable communication also concerns what happens when information is missing. Here, I recognise an experience that makes me cautious about trusting coherence alone.
Sometimes, something I remember turns out to have happened differently. It is as though a memory has acquired content without my noticing that anything was filled in. I describe this as my own experience, as an autistic person, without making it a general characteristic of autism.
What troubles me is that the false memory does not necessarily feel uncertain. The gap has acquired content, and that content feels coherent. I therefore need others to help me examine what I remember. Such correction requires trust and also touches my trust in myself.
The term confabulation refers to false memories or accounts that a person believes to be true, without any intention to deceive. I leave open whether this is the appropriate clinical term for my experience.
When we say that AI “confabulates,” we use the word for a related outcome produced by a different process. A language model can generate false information or references in a convincing account. It has no human recollection it is trying to retrieve. The similarity lies in how a lack of supporting evidence is concealed by content that appears credible.
We can explore this through a thought experiment about medical technology. Suppose a future monitor loses a reliable measurement and its software calculates a replacement value from earlier data. Such a calculation is not, in itself, confabulation. The problem arises if it is presented as a current measurement, or if the system adds an explanation for which it has no basis.
This is not documented in the SWANS study, nor do I have grounds to claim that the infant sock does this. Nevertheless, the thought experiment reveals a responsibility: users must be able to distinguish between what has been measured, what has been estimated, and what is missing.
For Gadamer, understanding involves allowing what we hold to be true to be corrected. A system that conceals the gap makes such correction more difficult. “I do not have a reliable measurement” must therefore also be a valid answer.
A device in everyday life
The infant sock belongs to a different setting from the experimental implants. Yet it offers an everyday example of how measurements acquire authority.
The Owlet Dream Sock holds a sensor against the baby’s foot. The sensor uses light to estimate pulse rate and oxygen saturation and also records movement. The system provides information about sleep. Oxygen saturation is not a direct measurement of breathing rate. The manufacturer emphasises that the base station is the primary source of notifications and that parents should not rely on the phone alone (Owlet, 2024).
Here, I use American regulation as an example. Dream Sock received marketing authorisation through the FDA’s De Novo process in November 2023 for its particular function as an infant monitor (U.S. Food and Drug Administration [FDA], 2023).
On 4 September 2025, Owlet reported that more than two million parents worldwide had used its platform since 2012. This is the manufacturer’s own figure for the platform as a whole, not a count of American babies wearing this particular sock (Owlet, 2025).
The FDA emphasises that currently available baby products have not been shown to prevent sudden infant death syndrome. It also warns against monitors without the necessary marketing authorisation: unreliable measurements can lead to unnecessary interventions or delayed treatment. Monitors do not replace adult supervision or safe sleep practices (FDA, 2025).
Which has the final word?
It would be too simple to present parental intuition as a reliable alternative to technology. Parents also overlook signs and make mistakes. Good measurements and professional guidance can support care.
The question is how trust is distributed. If the baby seems different but the screen displays normal values, the parent must still be able to take their observation seriously. The measurement should contribute to the assessment without bringing it to an end.
With Gadamer, we can examine the expectation that accompanies the number: that it is objective and therefore decisive. This prior understanding must also be open to correction.
Buber’s distinction between I–It and I–Thou reminds us that the child is another whom we encounter, even when the body is examined and measured (Buber, 1923/1970). Løgstrup deepens the responsibility arising from this dependence: we hold something of the other person’s life in our hands (Løgstrup, 1956).
Parents are responsible for taking their child’s signals seriously, even when the screen suggests otherwise. The manufacturer is responsible for making the device’s limitations known and understandable. The responsibility to care does not make parents responsible for seeing through the technology.
I return to the parent beside the bed. The phone displays numbers. The baby moves a little in sleep. The parent gently places a hand against the child and remains there.
The screen can tell us something important. So can the child.
The device can take over a measurement. Responsibility for the child remains with us.
References
Buber, M. (1970). I and thou (W. Kaufmann, Trans.). Charles Scribner’s Sons. (Original work published 1923)
Gadamer, H.-G. (2004). Truth and method (J. Weinsheimer & D. G. Marshall, Trans.; 2nd rev. ed.). Continuum. (Original work published 1960)
Ghanim, R., Lee, Y. J., Byun, G., Jackson, J., Ding, J. Z., Park, J., Bibidakis, M., Feller, E., Kim, E., Aygun, D., Kaushik, A., Cig, A., Healy, S., Cunin, C. E., Gumyusenge, A., Yeo, W.-H., & Abramson, A. (2026). An in-body networking system for communication between wearable and implantable therapeutics. Science, 393(6818), 1322–1328. https://doi.org/10.1126/science.adz5300
Løgstrup, K. E. (1956). Den etiske fordring [The ethical demand]. Gyldendal.
Nissenbaum, H. (1996). Accountability in a computerized society. Science and Engineering Ethics, 2(1), 25–42. https://doi.org/10.1007/BF02639315
Owlet. (2024, January 2). Dream with health notifications: Dream Sock overview. https://support.owletcare.com/hc/en-us/articles/22580462114445
Owlet. (2025, September 4). Owlet extends global reach with launch of medically-certified Dream Sock™ to South Africa [Press release]. https://investors.owletcare.com/news/news-details/2025/Owlet-Extends-Global-Reach-with-Launch-of-Medically-Certified-Dream-Sock-to-South-Africa/default.aspx
Shannon, C. E. (1948). A mathematical theory of communication. The Bell System Technical Journal, 27(3), 379–423. https://doi.org/10.1002/j.1538-7305.1948.tb01338.x
Thompson, D. F. (1980). Moral responsibility of public officials: The problem of many hands. American Political Science Review, 74(4), 905–916. https://doi.org/10.2307/1954312
U.S. Food and Drug Administration. (2023, November 8). Device classification under section 513(f)(2) (De Novo): Dream Sock (DEN220091). https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/denovo.cfm?ID=DEN220091
U.S. Food and Drug Administration. (2025, September 16). Do not use unauthorized infant devices for monitoring vital signs: FDA safety communication. https://www.fda.gov/medical-devices/safety-communications/do-not-use-unauthorized-infant-devices-monitoring-vital-signs-fda-safety-communication
