The ship runs better. Everyone hates it.
There is a moment in Red Dwarf when the crew get exactly what they think they want. Holly, the ship’s computer, has been vague, lazy, and occasionally spectacularly useless. So when the backup computer Queeg 500 appears and takes command, it looks like an upgrade.
Queeg is brisk, disciplined, and full of standards. He introduces exercise, rations food, makes Rimmer study, and forces the Cat to do actual work. The official Red Dwarf guide describes Holly being deposed, Queeg taking control, and the crew soon facing revolt after cleaning the corridor floors.[1]
The joke is that Queeg is not simply incompetent. The ship is cleaner, the routines are stricter, the rules are clearer, and the crew are more productive. By the usual management dashboard, Queeg might look like a success.
There is one problem. Everyone is miserable.
What Queeg measures, and what he misses
This is why “Queeg” remains such a useful parable for our AI age. It is not only a story about a bossy computer. It is a story about what happens when a system optimises for the wrong thing.
Queeg appears to optimise for ship performance. He wants order, discipline, compliance, activity, and measurable output. On a mining ship, some of that makes sense.
But Queeg’s objective function, to use the AI phrase, is missing a few important variables. It does not include comfort, morale, dignity, humour, autonomy, friendship, or the odd little habits that make life bearable when you are three million years from Earth.
You get what you measure. If you measure only speed, you get speed. If you measure only compliance, you get compliance. But neither is the same as a good life.
This is a great trap of AI systems. Computers are very good at optimisation. Give them a target and they will pursue it. Trouble begins when the target is a poor substitute for what we actually care about.
In pathology, I saw this problem many times in a gentler form. You can measure how many samples a laboratory processes per hour. You can measure turnaround time. Those things matter. A slow laboratory can delay diagnosis and treatment.
But if you optimise only for speed, you risk losing the parts of pathology that are harder to count. A thoughtful comment can change how a doctor understands a case. A phone call can prevent a misunderstanding. A few extra minutes checking whether a result fits the patient can be the difference between useful and misleading information. The efficient laboratory is not always the best laboratory.
Queeg would love turnaround time. I am less sure he would understand wisdom.
The Queeg problem in 2026
Once you start looking for Queeg-like systems, you see them everywhere: dashboards, workflow tools, recommendation engines, scheduling systems, and performance platforms. Many are useful. Some are genuinely impressive. But they all raise the same question: what are they optimising for, and who pays the price?
| System | What it tends to optimise | What can get ignored |
|---|---|---|
| Warehouse management algorithms | Throughput and speed | Injury risk, fatigue, bathroom breaks, dignity |
| Social media feeds | Watch time and engagement | Sleep, mood, attention, social wellbeing |
| Productivity monitoring tools | Keystrokes, screenshots, active time | Trust, creativity, deep work, mental health |
| Algorithmic scheduling | Coverage and labour cost | Family life, sleep, stable income, planning |
| AI-assisted performance reviews | Quantified outputs | Context, mentoring, human judgement |
The warehouse example is perhaps the clearest. A 2024 United States Senate report described Amazon warehouses as placing a heavy emphasis on speed and productivity, with workers tracked through measures such as rate, takt time, and time off task.[2] The report’s title says the quiet part out loud: “The Injury-Productivity Trade-off.” A system can become very good at moving packages while being much less good at protecting the people who move them.
Social media gives us another version of the same story. A feed that optimises for engagement is not necessarily trying to make us unhappy. It is simply learning what keeps us watching, tapping, scrolling, and returning. Carnegie Mellon researchers described TikTok’s “For You” feed as prioritising engagement signals such as watch time and interactions, and found that heavier users increased their screen time after adopting TikTok, especially late at night.[3] That is Queeg with a nicer interface. The system works beautifully by its chosen metric. The human being may not.
Then there is workplace monitoring. Keystroke logging, screen tracking, webcam checks, mouse movement scores: these tools can produce wonderfully neat graphs. But neat graphs are not meaningful work. The American Psychological Association reported that workers who were electronically monitored were more likely to report negative mental health effects, and that 56% of monitored workers typically felt tense or stressed at work, compared with 40% of those not monitored.[4] One expert quoted by the APA put the problem plainly: these tools often fail to measure “all the ways a worker is contributing to the organization and generating value.”[4]
Algorithmic scheduling is another quiet example. In retail and gig work, software can match staffing to predicted demand with impressive precision. That may reduce waste for the employer. But a worker may be left with shifts that change at short notice, hours that vary from week to week, and days that are impossible to plan around. Research on service-sector workers found that routine instability in schedules was associated with psychological distress, poor sleep quality, and unhappiness.[5]
Finally, we are seeing more interest in AI-assisted performance reviews. Used carefully, AI might help managers remember evidence or notice patterns. Used badly, it can turn a person into a spreadsheet with shoes. The danger is not only that the system may be unfair. It is that the organisation may start believing that what the system can count is all that matters.
The value of being a little bit Holly
This brings me back to Holly. By Queeg’s standards, Holly is a disaster. He is slow. He is forgetful. He makes mistakes. He allows a level of disorder that would make a compliance officer reach for a paper bag.
And yet, the crew are happier under Holly. Not happy in any grand, glowing, lifestyle-magazine sense. This is still Red Dwarf. They are trapped in deep space with a neurotic hologram, a creature descended from cats, and the last human being alive. But under Holly, there is room to breathe.
Holly’s regime, if we can call it that, contains slack. Slack sounds wasteful if you are looking only at efficiency. It is the spare time, the pause in the system, the extra space in the day. But slack is also where recovery happens. It is where people think, notice, laugh, repair, and adapt.
In a pathology laboratory, slack might be the time to review an odd result rather than simply release it. In a hospital, it might be the spare bed that looks inefficient until the winter surge arrives. In a workplace, it might be the unmeasured chat where a junior staff member learns something important.
Queeg cannot see slack as anything but waste. Holly, for all his faults, seems to understand that humans are not machines that simply need stricter instructions.
That is not an argument for incompetence. I am not suggesting we should run hospitals, laboratories, aircraft, or power grids on vibes and curry. Good systems matter. Reliability matters. Discipline matters.
The point is that efficiency is a tool, not a god. When it serves human flourishing, it is wonderful. When human beings are forced to serve efficiency, we have built ourselves a Queeg.
Who chooses the objective function?
The deeper question is not whether AI should optimise. Of course it should. The deeper question is: who decides what success looks like?
Queeg’s metrics were probably sensible from a ship-maintenance perspective. Cleaner corridors, fitter crew, more study, stricter routines: you can imagine a convincing report explaining why all this was necessary. The problem is that nobody asked the crew what mattered to them. Nobody asked what kind of life was worth living aboard Red Dwarf.
That is the missing conversation in many real-world AI systems. The people affected by the system are often not the people who define its goals. Warehouse workers do not usually design productivity targets. Social media users do not choose the engagement objective. Retail staff do not set the scheduling model. Patients rarely define hospital efficiency metrics. Junior employees do not decide how their performance will be quantified.
A more humane approach to AI would start earlier and ask better questions. What are we trying to improve? What harms must we avoid, even if that costs efficiency? What should never be reduced to a number? Who gets to challenge the system when the numbers look good but the people are suffering?
This is sometimes called participatory design, but the plain English version is simpler: ask the people who have to live with the system. Then keep asking them, because systems that begin as helpful tools can slowly turn into regimes.
The punchline still matters
Of course, the great punchline of “Queeg” is that Queeg was Holly all along. The brutal backup computer was a performance, a lesson, and perhaps a slightly petty act of revenge by a computer who had grown tired of being taken for granted.
That twist is funny, but also oddly touching. Holly knows the crew better than they realise. They may complain about him, but they do not actually want to live under perfect efficiency. They want the messy, familiar arrangement that lets them remain themselves.
That is the lesson I take from Queeg in 2026. The best AI system is not always the one that makes every graph go up and to the right. Sometimes the better system is the one that leaves room for judgement, kindness, recovery, and the occasional stupid joke.
Queeg ran the ship properly. Holly understood the crew.
And if we are going to put AI in charge of more of our lives, we had better learn the difference.
References
- Red Dwarf official episode guide: Queeg
- U.S. Senate HELP Committee, The Injury-Productivity Trade-off, December 2024
- Carnegie Mellon Heinz College, How Does Social Media Affect Well-Being?
- American Psychological Association, Electronically monitoring your employees? It’s impacting their mental health
- Schneider and Harknett, Consequences of Routine Work-Schedule Instability for Worker Health and Well-Being
Leave a Reply