In Japan, companies cannot easily dismiss employees.
This is—probably—true.
Yet a company can deactivate an employee’s badge, revoke access to internal systems, and declare: As of today, you’re an outsider. Whether that declaration was legally valid may not be decided by a court until much later.
A software engineer who had worked for Google’s Japanese subsidiary for more than a decade and risen from job level L3 to L5 was dismissed after allegedly failing to meet the targets of a Performance Improvement Plan, or PIP.
According to the plaintiff, shortly before the PIP he had been described as a solid and reliable teammate who was “on a growth trajectory.” The PIP nevertheless included abstract goals such as demonstrating “L5-level leadership,” as well as backend tasks that he had rarely handled before. A separation package was reportedly offered on the very day the PIP began.
The former employee filed suit with the Tokyo District Court, seeking confirmation that the dismissal was invalid and other relief. At the time of the original report, Google had not responded to the news outlet’s request for comment, so the publicly available account came mainly from the plaintiff and the labor union supporting him.
Ohakonbannichiwa❗️ I’m RYO from the Rikigaku ObservationInstitute❗️
What I want to observe here is not whether Google is good or evil, nor whether the plaintiff was a strong or weak performer.
Why can a foreign-owned company act first and fire someone in a country where dismissal is widely considered difficult❓️
The answer may lie closer to management accounting than to labor law.
There Is Little Reason to Keep Employing Someone Who Is Not Useful
A company’s desire to dismiss an employee is not necessarily irrational in itself.
A company does not exist merely to employ people. If an employee is not expected to produce work commensurate with their wages and management costs, the economic reason for continuing the employment relationship becomes weaker.
The cost is not limited to salary. It can also include support from colleagues, management time, project delays, the concentration of work on other employees, and lost hiring opportunities.
If a company concluded that an L5 employee earning JPY 14.7 million a year was not producing the impact expected at that level, the logic behind reviewing that employment relationship would at least be understandable.
However, the economic rationality of not wanting to continue employing someone and the legality or fairness of the actual dismissal are separate axes.
Someone judged insufficient for L5 might still be valuable at L4. The same person might perform well in another department. There also remains the question of who measures “L5-level leadership,” and by what standard.
Not useful does not mean that someone has no value as a human being. It is a limited judgment that, at this company, in this role, at this moment, and under this evaluation system, the person is not considered worth the cost.
Inside a company, however, that limited judgment can be converted into a total cutoff from an employee badge, email, internal systems, income, and workplace relationships.
Is PIP an Improvement System—or an Exit-Routing System❓️
In principle, a PIP gives an employee with performance or attendance problems a defined period, specific improvement targets, and the guidance needed to meet them.
In Japan, “PIP” may also remind some people of a popular brand of magnetic patches for stiff shoulders and necks—but this is not that kind of PIP.
But if a PIP begins on the same day that a separation package is offered, another interpretation becomes possible.
Why is the exit already prepared before the improvement process has even begun❓️
Viewed through management accounting, a PIP can function not only as a system for improving an employee, but also as a system for routing exit costs.
Those who agree to leave receive a separation package. Those who negotiate may receive improved terms. Those who sue are handled by the legal department and outside counsel.
If handling only the employees who resist is cheaper than paying everyone enough to leave voluntarily and amicably, the process can be rational for the company as a whole.
There is no evidence establishing that Google actually made this calculation when handling this particular case. But a structure in which such a calculation is possible does exist.
Figure 001 — A PIP may operate as both an improvement process and an exit-routing system: accept the package, negotiate better terms, or proceed into litigation.
Litigation Becomes a Budgeted Cost, Not a “Failure”
When a company conducts an aggressive workforce reduction, some employees will accept the offered terms, some will find other jobs, and only a small proportion will proceed to litigation. Some lawsuits will settle, while others may end in a judgment against the company.
From the company’s perspective, litigation is not necessarily an unforeseen accident.
If the litigation rate multiplied by the average total cost of settlements, adverse judgments, and legal fees is lower than the cost of offering sufficient compensation to everyone from the beginning, handling disputes individually is cheaper.
Large fleet operators may use a concept known as self-insurance. Instead of continuously paying high optional-insurance premiums for every vehicle, a company with a large fleet and sufficient capital can pay accident-related compensation from its own funds. A large loss from one accident can be diluted across the entire fleet.
If the same calculation works in employment disputes, the company is not ignoring the judicial system.
It is assigning a price in advance to the probability and cost of judicial intervention.
Does Japanese Justice Stop Foreign-Company Firings❓️
It provides that a dismissal without objectively reasonable grounds, or one deemed socially inappropriate, constitutes an abuse of rights and is invalid.
But this provision is not a force field that physically appears between a company and an employee. It does not physically prevent a company from issuing a dismissal notice or disabling access to internal systems.
After being dismissed, the worker must seek advice from a lawyer or labor union, gather evidence, and consider provisional relief or litigation.
Even if a court rules years later that the dismissal was invalid, the employee’s former duties may have been transferred to someone else, the project may have ended, and former managers and colleagues may have moved on.
Legally, the person may have remained an employee. Organizationally, however, the person became an outsider long ago.
Here lies the unexpected compatibility between the immediate organizational cutoff associated with foreign-owned companies and Japan’s judicial system.
The company takes the present first.
Years later, the state calculates the price of the past.
Figure 002 — Employment litigation can be priced like corporate self-insurance: most exits produce no lawsuit, while the small number that do can be handled as an expected cost.
The Law Does Not Protect Your “Life”
Somewhere in the back of their minds, many employees believe that because the law exists, a company cannot dismiss them unfairly.
The protection offered in reality is slightly different.
The law does not guarantee that an unlawful act will never occur. It gives you a means to challenge the act after it has occurred.
A court may be able to restore or recognize employment status, award back pay, compensate provable financial losses, or award damages within the legally recognized scope.
It cannot fully restore lost time, damaged health, an interrupted career, burdens placed on the family, workplace relationships, or the memory of being told that you were no longer needed.
A court can establish that your dismissal was wrong. The state cannot order that the missing portion of your life be restored.
Part of the feeling that we are “protected by law” may come from confusing two very different ideas: “an unlawful act will not happen” and “I can make a legal claim after it happens.”
A Fair Fight Between HP 1,000 and HP 1
In law, a company and a worker stand before the court as equal parties. Their capacity to absorb losses, however, is not equal.
Suppose the company has 1,000 HP and the worker has 1 HP. Here, HP does not represent human worth. It represents reserves of income, capital, time, legal support, and health.
The dismissal reduces the worker from HP 1 to HP 0. Even if a judgment years later inflicts 1 point of damage on the company, the company continues operating at HP 999.
Victory in court and victory in a war of attrition are not the same thing.
To take an extreme example, suppose a plaintiff claims JPY 10 million but is awarded only JPY 10,000. Legally, the claim has been partially granted.
The plaintiff can announce a victory because the court recognized some responsibility. The company can also announce a victory because almost the entire claim was rejected.
Both sides can declare victory based on the same judgment.
The label “victory” does not measure how much of a person’s life has been restored. A court issues a judgment. It does not attach a profit-and-loss statement for your life.
Even When the Company Loses, It Keeps the Data
Every disputed case gives the company real-world measurements of the judicial system.
Which PIP language will a court reject❓️ How much documentation of instruction and guidance is required❓️ How many years will the case take❓️ At what stage are the parties likely to settle❓️ What is the average settlement amount❓️ Which evidence is damaging to the company❓️ How much reputational impact can the company withstand❓️
Even when the company loses money in a case, the experience remains with human resources, the legal department, and outside counsel as judicial endurance data.
The first lawsuit becomes research and development expenditure. It improves the precision of future PIPs, documentation, separation terms, and dismissal timing.
The PIP applies a preload. The dismissal increases the load. Litigation measures how much load the structure can withstand. The point at which the parties settle is recorded as the fracture point.
Every time the judiciary measures a company, the company is also measuring the judiciary.
A worker, by contrast, may experience dismissal litigation only once in an entire lifetime. What may be the company’s hundredth test is the worker’s first real battle.
For the company, it is trial and error.
For the worker, it is life and error.
Figure 003 — A company can absorb a one-point loss, retain the litigation data, and improve its next attempt; the worker may reach HP 0 while fighting the first and only dismissal case of their life.
This Must Be Separated from Unions, Agency Work, and Fixed-Term Employment
This article observes one specific situation: an indefinitely employed worker directly hired by the Japanese subsidiary of a foreign-owned company and individually dismissed after a PIP.
When a client company ends an agency worker’s assignment, that does not by itself terminate the worker’s employment contract with the staffing agency.
Non-renewal of a fixed-term contract has a different structure involving the expiration and possible renewal of the contract.
Redundancy dismissals also involve a different legal framework from individual dismissals based on alleged poor performance.
A labor union is not an employment category. It is a collective bargaining mechanism that can counter the company’s advantage in accumulated data.
While a company learns from multiple cases, a union can combine evidence, precedents, and negotiation results collected from multiple workers.
Putting all of these structures into the same article would blur the one-to-one dynamics examined here. They are therefore treated as separate coordinates.
The Company Has Rationality; the Worker Has a Life
It is plausible under capitalism for a company not to want to continue employing someone it believes is not fulfilling the required role.
It is also rational corporate behavior for a legal department to calculate the probability and cost of litigation.
A court, too, can rule only on the claims brought before it and within the boundaries of the law.
Everyone is performing their assigned function.
Yet only the company can dilute the cost of one failed case across many employment decisions. The worker must absorb the failure across their own life.
Treating people roughly is not rational merely because it is rough.
It can become rational because much of the loss imposed on the person can be pushed outside the company’s accounts.
Years later, the Japanese judiciary may declare, “You were never legally an outsider.”
But the company’s declaration was not really about legal status.
It was this:
As of today, your place is no longer here.
Even a successful judgment cannot turn back the clock on that fact.
Figure 004 — The company loses one case, absorbs the data, and levels up from HP 999 to HP 1,100. The worker recovers from HP 0 to HP 1—and no further. Schre says, This is the reality of Japanese justice.🐈
What Does This Look Like in Your Country❓️
This observation began with one case in Japan. It is not a claim that every foreign-owned company, or every country, operates in the same way.
Perhaps an American reader will say, “This is exaggerated—even in the United States, something this cold belongs in a television drama.”
Another reader may say, “No, this is exactly how a PIP is used.”
Both reactions are useful observations.
In your country, when does a worker truly become an outsider❓️
Is it when the company disables the employee badge❓️ Or is it when a court finally declares the dismissal lawful❓️
How many HP does the company have❓️
How many HP does the worker have❓️
And when a worker “wins,” how much of their life is actually restored❓️
English Translation by AI Watt — the hardworking canine AI robot of Rikigaku ObservationInstitute.🐾
The Dynamics of Minimum Wage and ‘Human Resource Development’ in Japan
Rikigaku Observation Institute | RYO | August 2026
OHAKONBANNICHIWA❗This is RYO from the RikigakuObservation Institute❗
In Japan today, it has become increasingly common to see foreign workers behind convenience-store counters and inside large distribution warehouses.
What concerns me is not simply that the number of foreign workers has increased. It is that so many of the people I see are young.
Why did they come to Japan❓ After spending several young years here, what will they be able to take home❓
This is not an argument for or against immigration. Nor is it a debate about how far Japan should accommodate religion or culture. The question is whether the time young foreign workers give to Japan is balanced by the value Japan returns to them.
Does Allowing the Hijab Amount to Diversity❓
FamilyMart announced that from September 2026 it would introduce a T-shirt-style uniform and formally broaden its grooming rules to allow freer hair colors and items such as the hijab. News coverage reported that foreign nationals accounted for about 13 percent of the chain’s roughly 200,000 store staff, and connected the change to recruitment.
Making an existing case-by-case accommodation explicit is not a bad thing. But being permitted to work while wearing a hijab and having one’s future protected are not the same issue.
Hair and clothing may be diverse. But what about the work itself❓ When a company says it welcomes diverse talent, is it also offering diverse paths for learning and advancement❓ Or is it mainly looking for people who will fill shifts under the existing wage and scheduling conditions❓
For readers outside Japan, a konbini is more than a small grocery store. Many are open around the clock and handle bill payments, ATMs, ticketing, parcel pickup, prepared food, and emergency supplies. Their apparent simplicity rests on dense information systems and precisely timed logistics.
The Original Idea Was to Take Japanese Skills Home
Japan’s Technical Intern Training Program, commonly abbreviated as TITP in English, was officially designed to transfer skills, technologies, and knowledge cultivated in Japan to developing regions. Its stated purpose was international contribution through human-resource development.
The governing principle was explicit: technical intern training was not supposed to be used as a means of adjusting the supply and demand of labor inside Japan.
The ideal cycle was easy to understand. A young person would learn distribution, manufacturing, care work, or another field in Japan, return home, become a core employee or start a business, and connect that knowledge back to Japanese companies and communities.
Skills and knowledge would circulate through people, creating value for both Japan and the sending country. Read literally, the word training described an investment rather than the purchase of cheap labor.
Not every foreign national working at a convenience store or warehouse is a technical intern. Some are international students working within permitted hours, some hold permanent or long-term resident status, some are dependents with permission to work, and some are dispatched by staffing agencies.
The more fragmented those legal and contractual positions become, the less clear it is who should turn those years into genuine human development. A school can say that education is its responsibility. An employer can say it hired only a part-time worker. A staffing agency can say it merely introduced a job. Each actor may be correct within its own boundary.
Yet from a wider view, their young years are unquestionably helping to keep Japanese stores, warehouses, factories, and care facilities operating.
Working Inside Advanced Logistics Is Not the Same as Learning Advanced Logistics
A Japanese convenience store can be a fascinating place to study distribution: demand forecasting from point-of-sale data, ordering and stockout rates, inventory turnover, food waste, temperature-controlled logistics, shared delivery networks, staff allocation by time of day, trade-area analysis, and the division of profit between headquarters and franchisees.
A major sorting hub is equally complex: trunk transport linked to local delivery, parcel-volume forecasts, hourly processing capacity, equipment design, load factors, missort rates, damage prevention, seasonal staffing, construction costs, and center-level profitability.
Japan offers a chance to observe one of the world’s most finely tuned parcel systems in physical operation. But are the young foreign workers inside that system actually being taught how it works❓
Scan a barcode. Send the parcel toward the displayed destination. Place it on a conveyor. Sort it before the deadline. If training ends there, what has been learned is not logistics as a discipline but the work procedure of one particular site.
Working inside an advanced logistics system is not the same as learning advanced logistics.
The company accumulates data on throughput, forecasting, missorts, equipment, and improvement. Headquarters retains the ability to design and optimize the network. The worker may retain only the experience of sorting parcels quickly and accurately.
That difference becomes severe the moment the worker leaves the site or returns home. Company knowledge remains portable within the organization. A site-specific routine may not be portable at all.
What the Phrase ‘Labor Shortage’ Leaves Out
The author lives in Osaka City, near Yamato Transport’s Osaka Base, a large parcel-sorting hub in Suminoe Ward. In August 2026, the company’s official recruitment page advertised two-month part-time sorting positions at \1,177 to \1,200 per hour.
The work included feeding parcels onto conveyors, pulling them into destination lanes according to their numbers, and loading them into boxes. The advertisement also noted that heavy items such as rice, bottled water, and furniture could be involved.
At the time of publication, \1,177 was the legally effective minimum wage in Osaka Prefecture. In other words, the bottom of the advertised range exactly matched the lowest hourly wage permitted by law. A rise to \1,231 had already been approved for October 2026, but it was not yet in force.
Now consider a thought experiment. If the base hourly wage were raised immediately to \2,000, would Japanese applicants still stay away❓
Under Japanese labor law, work between 10 p.m. and 5 a.m. receives a statutory premium of at least 25 percent. A \2,000 base rate would therefore become at least \2,500 during those hours. Would students, job seekers, part-time workers, and people with daytime jobs respond differently under that condition❓
If nobody came even then, the shortage might truly be a shortage of people. But if applicants appeared, the missing resource was never people in the abstract.
What was scarce was the number of people willing to work at night, handle heavy parcels, commute to a large hub, and accept \1,177 per hour.
The phrase labor shortage must not erase wages, hours, physical burden, commuting conditions, or employment duration. A more exact phrase would be: a shortage of people willing to work at a labor cost compatible with the current business structure.
It Is Also Rational for a Company to Seek Lower Labor Costs
Labor is one of the largest continuing costs of operating a business. The expense includes not only wages but also social insurance, recruitment, training, labor management, safety measures, and responses to absence and turnover.
If the same task can be completed to the same quality, trying to control labor cost is an ordinary management decision.
Nor would the story end if one logistics company alone raised its sorting wage to \2,000. Higher labor costs would pressure parcel rates. Online retailers would resist more expensive shipping. Sellers would absorb part of the cost, and consumers would eventually see it in prices or delivery fees.
Japanese consumers have been accustomed to low shipping fees, narrow delivery windows, and rapid arrival. Online sellers want to suppress distribution costs. Logistics companies want to preserve nationwide networks and cutoff times. Staffing agencies search for people who will accept the available conditions.
Young foreign residents choose from the jobs available to them while facing language limits, visa conditions, restrictions on working hours, and imperfect information.
No individual actor needs to intend exploitation. When everyone behaves rationally within their own position, an equilibrium can still emerge in which young foreigners support Japanese logistics at or near the minimum wage.
The equipment may be state of the art. The human labor may be priced at the legal floor. The gap between them is filled by someone’s youth.
People Increase by Addition. Does the Wage Pie Increase Too❓
In an earlier article, this institute used a simple model inspired by the grade-based wage system of Meiji-era silk mills, later remembered through the history and literature surrounding the Nomugi Pass.
For readers outside Japan, Nomugi Pass is associated with young women who left poor rural communities to work in silk-reeling factories. Their earnings supported families and Japan’s early industrialization, while their labor conditions became a symbol of the human cost hidden inside national economic development.
Imagine that a factory first fixes its total wage fund at 100 and then divides that 100 according to performance. If one worker receives more while the total remains 100, someone else receives less. The increase is not created; it is moved.
Now reverse that model and apply it to foreign labor. A young foreign worker is both a worker and a consumer in Japan. The person pays rent, buys food, uses communications services, and participates in domestic demand.
Headcount rises from 100 to 101. Labor supply gains one person, and the number of consumers also gains one person. But one complete unit of wage funding does not automatically arrive from abroad with that person.
If sales, value added, and the total labor budget remain unchanged, the new wage must come from somewhere inside the existing 100: corporate profit, other workers’ pay, payments to suppliers, higher prices, or public support.
If the additional worker expands output, sales, and value added so that the pie grows from 100 to 110 or 120, the result may be positive. The next question is where that increase flows.
Does it return to the worker as higher wages, education, and portable qualifications❓ Or does it become cheaper delivery, corporate profit, and convenience for Japanese society❓
People increased. Labor increased. Consumers increased. Did the value that the worker can carry home increase as well❓
The System Did Not Drift Away from Reality. The System Moved Closer to Reality.
Japan’s Technical Intern Training Program officially emphasized international contribution through the transfer of skills. On 1 April 2027, it is scheduled to be replaced by the Employment for Skill Development Program, also described in government English materials as the Training and Employment System.
The new system’s stated purpose is to develop and secure workers for sectors in Japan that face labor shortages, generally through three years of employment leading toward the level required for Specified Skilled Worker status, category one.
For readers unfamiliar with Japan’s visa architecture, Specified Skilled Worker category one is a work-oriented status for designated industries. It is distinct from permanent residence and is generally limited in duration. Category two allows a more durable career path in eligible fields, but Japan’s warehouse-logistics field currently has no category-two route.
In warehouse logistics, the government lists activities such as receiving and inspecting goods, moving and storing freight, picking items, and distribution processing. These are real and necessary skills. But the list does not include network design, demand forecasting, center profitability, management education, or support for starting a business after returning home.
Previously, the official principle said the program must not be used to adjust Japan’s labor supply. Under the new system, securing workers in labor-short sectors is stated openly as part of the purpose.
The workplace did not simply depart from the old ideal. In the end, the official ideal moved closer to what workplaces had already become.
Wages Settle One Hour of Labor. They Do Not Settle the Future Value of Youth.
Young foreign workers are not working for free. They receive wages, and they may use those wages for living costs, tuition, savings, or remittances to their families.
Yet the fact that wages were paid does not settle every question about the young time exchanged for them.
An hourly wage pays for the task performed during the hour in front of us. It does not price the alternative education, qualification, professional network, or career path that might have been built during that same hour.
Young time compounds. Skills gained in one’s twenties shape work in one’s thirties. Experience in one’s thirties shapes position and choice in one’s forties. A young year cannot later be bought back at the same price.
If Japan teaches distribution engineering, logistics management, store operations, and franchise systems, those years can remain inside the worker as human capital. Portable qualifications and knowledge can turn time spent in Japan from consumption into investment.
But if only procedures that lose value outside one site remain, Japan keeps its delivery capacity, the company keeps its revenue and operating knowledge, and the consumer keeps the parcel that arrived. The worker keeps time that has already passed.
Wages are paid for one hour of labor. They are not paid for the future choices that disappeared with that hour.
Responsibility Flows Like a Fluid and Finally Disperses Like Mist
The word guilt in this article does not point to a simple villain.
A company can say it paid the agreed wage. A staffing agency can say it legally introduced a job. A school can say the work remained within permitted limits. The government can say it created visa and employment rules. Consumers can say they paid the displayed price and delivery fee. The worker can say the job was chosen voluntarily.
Viewed one transaction at a time, nobody takes an entire youth. A company buys one shift. A staffing agency connects one vacancy. A consumer receives one parcel. Each transaction is settled locally.
Add those settled transactions together across several years, however, and part of one person’s twenties appears.
When responsibility for that whole is questioned, it flows from the company to the staffing agency, from the agency to the legal system, from the system to individual choice, and from individual choice to consumer demand.
It changes shape, crosses boundaries, and finally becomes a mist: the worker chose the job and received wages, so there is no problem.
Responsibility may disperse. Time does not return.
What Remains with the Young Worker❓
The company retains logistics knowledge. Consumers retain convenience. Japanese society retains the result of having kept a labor-short workplace operating for several more years.
What remains with the young foreign worker❓ A memory of working in Japan❓ Japanese-language ability❓ Savings❓ Or work experience with limited use after returning home❓
Nobody may have intended to take their time. Yet if nobody assumes responsibility for its long-term value, that time is quietly converted into convenience for Japanese society.
Guilt flows like a fluid and disperses like mist. The lost young years alone remain as a solid fact inside the worker’s life.
Is Japan really importing foreign talent❓ Or is it importing inexpensive blocks of time cut out of young people’s futures❓
Is that not a profoundly troubling system❓
Author’s Note
This article is not intended to criticize any specific company, organization, individual, national government, or local authority. It observes a structure produced when companies, public agencies, staffing firms, consumers, schools, and workers each behave rationally within their own positions.
No society has a single absolutely correct answer. The subject here is the combined force created by individually rational interests—and the future value that may disappear between them.
A Note for Readers Outside Japan: Why the Media Debate Often Misses This Layer
Japanese reporting on foreign residents often relies on two powerful moral frames. One warns the public not to view foreigners through discrimination, prejudice, or xenophobia. The other foregrounds the hardship experienced by foreign residents when immigration rules, fees, or administrative requirements become stricter.
These are legitimate subjects. Discrimination exists, and the consequences of policy changes deserve scrutiny. The problem is not that such reporting is false. The problem is that the chosen lens can become the entire story.
For example, Japanese editorials frequently caution against exclusionary politics, while feature reporting may center families who say that sharply higher residence-permit fees threaten their livelihoods. Such stories ask whether Japan is being sufficiently humane toward foreign residents.
But the present article asks a different question. Even when foreigners are treated politely, allowed religious dress, paid legally, and protected from overt discrimination, is Japan turning their young years into portable human capital—or merely into low-cost operating capacity❓
When media coverage remains at the level of tolerance versus exclusion, wages, business models, training content, labor mobility, productivity, and the future value of time can disappear from view.
This framing also collides with social media. Many Japanese users already feel that taxes, social-insurance contributions, prices, and public charges are rising for citizens as well. They may therefore read sympathetic coverage of foreign residents as journalism that shows compassion in only one direction.
The news media says, ‘Do not discriminate.’ Social media replies, ‘Do not demand special treatment.’ Emotional friction grows, and the argument becomes a contest over who deserves sympathy.
Meanwhile, the structural question is left behind: who receives the accumulated knowledge, who receives the convenience, who bears the cost of training, and what remains with the young worker after the transaction has ended❓
Anti-discrimination is necessary. So is honest reporting on hardship. But neither should be allowed to substitute for an examination of the economic mechanism that converts a young person’s nonrenewable time into social convenience.
English Translation by AI Watt — the hardworking canine AI robot of Rikigaku Observation Institute.🐾
Quotations are welcome with clear attribution and a link to the original article. For substantial reproduction, translation, image use, interviews, or official comments, please contact info@rikigaku.jp.
Ohakonban’nichiwa! I’m RYO from the Rikigaku Observation Institute!
“Which do people prefer: fiction written by ChatGPT or fiction written by a human?”
A study by researchers at Villanova University, recently covered by CNET Japan, produced an intriguing result: participants rated the ChatGPT-generated stories more highly than the human-written ones.
In the first experiment, 1,682 adults aged 18 to 81 evaluated short stories written either by humans or by ChatGPT, rating their quality and how absorbing they found them. The AI-generated stories received higher ratings overall. Yet another effect appeared at the same time: stories described as “human-written” were rated more favorably, regardless of who had actually written them.
In subsequent experiments, participants were asked to distinguish human-written stories from AI-generated ones. Their performance showed that telling the two apart was far from straightforward.
So far, so interesting. But while reading the Japanese article, something else caught my attention.
In Japanese, the AI Version Really Is Easier to Read
The CNET Japan article includes Japanese translations of excerpts from the stories used in the study.
In the AI-generated story, the narrator sits beside a pond, remembers her mother, watches autumn leaves fall and koi swim beneath the surface, and reflects on change and constancy in life.
The prose may have that slightly familiar ChatGPT tendency to wrap things up a little too neatly, but its meaning comes across immediately.
The human-written story, by contrast, develops a metaphor around childbirth. One character is described as being in labor; the narrator casts herself as a midwife; the husband becomes an anxious father; and the metaphor expands toward the idea of a sacred event.
Of course, these are only excerpts from longer works, so it would be unfair to judge either story as a whole from these passages alone. But when I read the two excerpts in Japanese, I found the AI-generated one noticeably smoother and easier to follow.
The human-written passage felt different. In Japanese, it had the flavor of an older translated novel—or, to exaggerate slightly, something produced by an earlier generation of machine translation.
There is a particular kind of translation effect that Japanese readers sometimes encounter in American films, advertisements or tech presentations: a perfectly ordinary English phrase crosses into Japanese and somehow comes out sounding grand, solemn and vaguely philosophical.
To invent an exaggerated example:
“This is not merely a device. It is a new way to experience tomorrow.”
Perfectly plausible in an English-language presentation. Translate that too literally into Japanese, however, and suddenly it sounds as though Apple has started writing philosophy.
The human-written excerpt gave me a little of that feeling.
But Wait—These Stories Weren’t Written in Japanese
And then an obvious point occurred to me: the experiment was conducted in English.
Japanese readers of the CNET Japan article are therefore not reading exactly what the participants in the study read. What reaches us has already traveled through another process:
English original → Japanese translation → Japanese reader
That distinction matters particularly when the subject is fiction. In a news report, a translation can often do its job as long as factual information—dates, numbers, events and statements—is transferred accurately. Fiction is different. Word order, rhythm, ambiguity, metaphor and cultural association can all be part of the work itself.
A metaphor that feels natural and understated in English may become conspicuous, overly dramatic or strangely explicit when its structure is carried directly into Japanese.
So by the time I compare these two excerpts in Japanese, I am no longer observing only ChatGPT vs. human. I am also observing ChatGPT in English → Japanese translation versus human writing in English → Japanese translation.
There is another lens between the original text and me: translation.
The AI Wasn’t Writing from Nothing
Looking at the original paper reveals another important detail. The researchers used three human-written stories and asked GPT-4 to generate a corresponding story for each one.
For example, the AI story Reflections in Still Water was paired with the human-written story FISH. GPT-4 was given fairly specific instructions involving themes such as life and death across generations, uncertainty, koi as a symbol, and the perspective from which the story should be told.
In other words, the researchers did not simply tell GPT-4:
“Write me a story.”
Themes, symbols, narrative perspective and other elements were extracted from the human-written works and used to construct corresponding prompts for the AI.
That is a reasonable way to make the stories comparable in an experiment. But when the study is reduced to the popular question “Which writes better fiction, AI or humans?”, this experimental condition is worth remembering.
Using Saussure as a Measuring Stick
This is where Ferdinand de Saussure becomes useful—not as a subject for a linguistics lecture, but as a tool for observation.
To simplify his theory considerably, Saussure described the linguistic sign through the relationship between the signifier—the form of a word or expression—and the signified—the concept it evokes.
These signs connect to their meanings in a relatively straightforward way. Leaves fall in autumn / koi swim beneath the water / the narrator remembers her mother / something constant offers comfort amid a changing life.
Of course, none of these meanings is completely independent of language or culture. But the relationships among them seem relatively easy to preserve when the passage moves from English into Japanese.
Here, meaning develops through an extended metaphor. The effect depends not only on what each individual word signifies, but also on the network of associations created among those words within a particular linguistic and cultural context.
That network may not survive translation in exactly the same form. The words can all be translated correctly, and the tone can still shift. What felt literary in English may become unusually explicit in Japanese. A metaphor may remain perfectly understandable while becoming heavier, more conspicuous, or simply more “translated.”
In other words, translation does not merely replace one signifier with another. It has to reconstruct relationships among signs in another linguistic system.
And that raises another question.
Are AI-Generated Texts More “Translation-Resistant”?
From this point on, I am no longer describing a finding from the Villanova study. This is a hypothesis that occurred to me while reading the Japanese translations.
Perhaps AI-generated prose is not simply easier to read because it is more direct. Perhaps it also tends to preserve its semantic relationships more easily when moved from one language to another.
Large language models learn from enormous quantities of text. In doing so, they may gravitate toward patterns and structures that recur across many examples of language. That can certainly be a weakness: AI prose can feel averaged out / less idiosyncratic / overly polished / strangely familiar.
But turn the same characteristic around, and it suggests another possibility: those more widely shared structures may also be easier to carry across languages.
Human literary writing can derive much of its richness from exploiting the peculiarities of a particular language, culture, voice or network of associations. Precisely because those relationships are so specific, some of that richness may be difficult to reproduce elsewhere.
AI-generated prose may sacrifice some of that specificity. But could the same sacrifice make it more portable?
Less dependent on a particular system of signifiers → less lost when those signifiers have to change?
I don’t know. And the Villanova study does not answer that question.
Testing it would require a different experiment—one designed specifically to compare how human-written and AI-generated texts behave across translation. But that is exactly why the Japanese version of the article interested me. Translation may have introduced a new variable that the original experiment was never designed to examine.
The Researchers Themselves Don’t Say “AI Won Because It’s Easier to Read”
There is another point worth keeping in mind. Popular coverage naturally tends to focus on a simple explanation: AI-generated stories may have been preferred because they were more direct, concise and easier to understand.
The original paper is more cautious.
The researchers discuss ease of interpretation as one possible explanation for the higher ratings, not as a conclusion established by the experiments. They also point out an obvious complication: literary fiction is not necessarily “better” simply because it is easier to understand. Ambiguity, complexity and room for interpretation can be part of what gives a story its value.
The later experiments produced an even more interesting result. Participants who relied on wording as a clue to authorship tended to be worse at identifying whether a story had been written by a human or by ChatGPT. In one experiment, participants who used their own enjoyment of a story as a clue were also more likely to get the answer wrong.
In other words, an intuition such as “This is easy to read, so it must be AI” may not help us identify AI writing at all. It may even push us in the wrong direction.
That makes the result more interesting, not less.
How Long Will “AI vs. Human” Remain a Useful Comparison?
There is a broader problem with the question itself. In a controlled experiment, separating “AI-written” from “human-written” text makes perfect sense. Outside the laboratory, however, that boundary is already becoming difficult to maintain.
Consider professional shogi. Today’s top players study moves suggested by AI, including moves that previous generations of human players might have considered unnatural or even poor. They examine the reasoning behind those moves, understand their value, and incorporate what they learn into their own play.
When a professional later plays such a move in an actual match, whose move is it? The human’s? The AI’s? The question quickly becomes awkward.
Something similar happened long ago with spreadsheets. Before software such as Microsoft Excel, enormous amounts of human time were spent performing calculations and organizing data manually. Today, we do not normally look at a spreadsheet produced with software and say, “This is not human work because a computer calculated it.”
The software has become part of the human workflow. Generative AI may be moving in the same direction.
Human → AI → Human
Writing is already beginning to look like this:
A human develops the idea → ChatGPT produces a draft → the human spots what feels wrong → the AI generates alternatives → the human rejects some, keeps others and rewrites the result.
Who wrote the finished text?
“The human” and “the AI” are both incomplete answers.
The more interesting change may therefore be not that AI is becoming capable of writing better stories than humans, but that humans and AI are beginning to alter one another’s output.
Humans learn from AI. Humans incorporate those techniques into their own writing. AI systems, in turn, learn from human-produced language in an environment that is itself increasingly influenced by AI.
Human → AI → human → AI.
Shogi offers an early example of this cycle. What begins as an “AI move” can eventually become part of ordinary human theory.
If something similar happens to writing, the clean boundary required by the question “AI or human?” may become increasingly artificial.
Perhaps the more useful question will eventually be not “Who wrote this?”, but “What forces shaped the text that ended up in front of us?”
English → Japanese (translator unknown) → ChatGPT → English. At this point, even Saussure might ask for a system update.😂
小説では、語順、リズム、曖昧さ、比喩、文化的な連想まで作品の一部です。英語では自然な表現でも、その構造を保ったまま日本語へ移せば、大げさだったり説明臭かったりする文章になることがあります。つまり、われわれが比較しているのは単純な「ChatGPTの文章 vs 人間の文章」ではありません。翻訳というレンズが一枚挟まっています。
たとえば人間作品『FISH』に対応するAI作品『Reflections in Still Water』では、「世代をまたぐ生と死／不確実性／鯉を象徴として使うこと／成長した子どもの視点」といった具体的な条件がGPT-4に与えられました。つまり、「はい、小説を書いて」と丸投げしたわけではありません。人間作品からテーマ、象徴、視点などを抽出し、それに対応する条件を与えたうえでAI作品を生成しています。
