Robotics Fundamentals
Robotics, Ethics and Society
Robots already work in factories, hospitals, farms and homes. Explore the real questions they raise about safety, accountability, work and fairness, beyond Asimov's three laws.
beginner · 12 min read
Lesson content
Robots are already here. Welding cars. Dispensing medication. Milking cows. Vacuuming living rooms. Inspecting sewers. Driving around on Mars.
That changes the nature of the ethical questions completely. They are not speculative any more.
Most of them turn out to be design questions with deadlines, things somebody has to decide before the next build, rather than philosophy.
Let us start with the famous answer, because it is the wrong one.
Asimov's Three Laws of Robotics: do not harm a human or through inaction allow a human to come to harm; obey orders unless they conflict with the first law; protect yourself unless that conflicts with the first two.
Elegant. Memorable. Completely unimplementable.
Why can't we just program Asimov's laws?
Because not one word in them means anything to a machine.
Start with "harm". A robot slams to a stop to avoid hitting someone and the person walking behind it falls over. Did it harm them? Which harm was it supposed to weigh against which?
"Or through inaction allow harm" is worse. Taken literally, it makes the robot responsible for every bad outcome anywhere in the world it could conceivably have prevented. That is not a hard computation. It is an impossible one.
And underneath both, the laws quietly assume the robot can reliably recognise humans and predict consequences. Those are precisely the capabilities we do not have.
Asimov knew all of this. Nearly every story he wrote about the laws is a story about them going wrong. They are a plot device and reading them as a specification is a category error.
What actually exists is far less romantic and far more useful: standards.
ISO 10218 governs industrial robot safety.
ISO/TS 15066 defines the four collaborative operating modes and publishes measured pain thresholds for every part of the body, so "safe contact force" is an actual number in newtons rather than a wish.
IEC 61508 governs functional safety.
Notice how safety is actually achieved: hardware limits, redundant sensing, risk assessment and testing. Never by a rule pasted on top of the software.
The four questions that actually come up
Safety: who can it hurt and how badly?
Answered by engineering. Force and speed limits, separation monitoring, emergency stops and a documented risk assessment for the specific application. Not by the robot's intentions, because it has none.
Accountability: who answers when it goes wrong?
The chain runs designer → manufacturer → integrator → operator. The robot is never a link in it, because it is not a legal person and cannot bear responsibility. Every autonomous decision must be traceable to a human decision about how the system should behave.
Work: whose job changes and who decides?
Robots reliably displace tasks and unevenly displace jobs. The tasks that go first are the dangerous, dull and dirty ones, which is good. But the person doing them does not automatically get the new job that replaces it.
Fairness: who gets the benefit and who bears the cost?
The gains from automation and the disruption from it usually land on different people. That is a distribution question, settled by policy rather than by engineering. Engineers are still in the room when the systems are designed.
| Claim about robots and jobs | What the evidence supports |
|---|---|
| "Robots will eliminate most jobs" | Not supported. Automation has consistently changed the mix of work more than the amount of it. |
| "Robots only create better jobs" | Also not supported. New jobs often need different skills, in different places, for different people. |
| "Robots take tasks, not jobs" | Well supported. Most occupations are bundles of tasks and automation takes some of them. |
| "The effect is uneven" | Strongly supported. Routine manual and routine cognitive work is affected first and hardest. |
| "Adjustment is automatic" | Not supported. Retraining, relocation and timing are real costs borne by real people. |
"Robots will eliminate most jobs"
- What the evidence supports
- Not supported. Automation has consistently changed the mix of work more than the amount of it.
"Robots only create better jobs"
- What the evidence supports
- Also not supported. New jobs often need different skills, in different places, for different people.
"Robots take tasks, not jobs"
- What the evidence supports
- Well supported. Most occupations are bundles of tasks and automation takes some of them.
"The effect is uneven"
- What the evidence supports
- Strongly supported. Routine manual and routine cognitive work is affected first and hardest.
"Adjustment is automatic"
- What the evidence supports
- Not supported. Retraining, relocation and timing are real costs borne by real people.
Example
Accountability in practice: a warehouse injury
A warehouse robot turns a corner and knocks over a worker. Who is responsible?
The designer chose the sensor's field of view and the stopping distance.
The manufacturer built it to that spec and certified it.
The integrator laid out the aisles and set the speed limit for this site.
The operator decided how much training staff got and whether the aisle stayed clear.
Any one of those four decisions could be the one that mattered. Investigating the accident means reconstructing all four.
Now notice what is not on that list: the robot's own choice. "The algorithm decided" is never an answer. It just tells you which human decision to go and look at.
Two more questions get sharper every year, as robots move out of controlled spaces.
Privacy. A robot that drives around a home or a hospital ward is a camera, a microphone and a map-maker on wheels. What it records, where that data goes and who can look at it are all decisions a person makes on purpose.
Deception. A robot that appears to understand, to care, or to feel is inviting a relationship it cannot actually hold up. That matters a great deal for children, for people with dementia and for anyone who is isolated.
Ethics is a design phase, not a review stage
By the time a robot is built, its speed, its mass, its sensor coverage and what it does when something fails are all fixed.
Those are exactly the properties that decide whether it is safe and whether anyone can be held to account for it.
So an ethics review at the end of a project cannot change anything. It can only write down what was already decided months ago.
These questions belong in the concept phase, sitting on the same page as payload and cycle time.
Treating "the algorithm decided" as an explanation.
Someone chose the algorithm, its training data, its thresholds and what it does when uncertain. Those are the decisions to examine.
Assuming safety standards make an application safe.
Standards certify the robot. The application (tool, workpiece, layout, speed, people) needs its own risk assessment, redone whenever any of those change.
Debating humanoid rights while ignoring warehouse injury rates.
The urgent ethical questions in robotics are about existing machines and the people working beside them today.
Key takeaway
- Robots already work across industry, medicine, farming, homes and space, so the questions are concrete.
- Asimov's Three Laws are fiction and his own stories are about them failing. Real safety comes from ISO 10218 and ISO/TS 15066.
- Accountability runs designer → manufacturer → integrator → operator and stops above the machine.
- Automation takes tasks unevenly and the gains and the disruption land on different people.
- Safety, accountability, privacy and honesty are design decisions, made early or not at all.
Finished this lesson?
Guest progress is saved on this device.