Introduction

Hello, this is Nishihara.

When you gather information, what sources do you rely on?

Ask an AI, and it organizes what you want to know in moments. Open a social feed, and you can see how the world is reacting right away. News arrives one breaking headline after another.

The speed at which we can collect information is far greater than it used to be.

But information arriving quickly and information being sufficiently verified are two different things. Early-stage information is often added to or corrected later. Even something shared widely is not necessarily settled fact.

In an earlier article, I wrote that in the age of AI, judgment and responsibility are what remain with humans.

Yet if the information that judgment rests on is wrong, then no matter how sound the logic, it becomes hard to properly evaluate whether the judgment itself was valid. Even when the outcome is good, it becomes difficult to tell whether the decision was actually right or whether it simply happened to work out.

The quality of a judgment is strongly shaped by the quality of the information behind it.

At the same time, what matters in gathering information is not only taking in what looks useful or hunting for high-accuracy sources. That effort matters, of course. But when everyone is aiming for the same thing, it becomes hard to create real difference or value from that alone.

What matters is how you read the information you obtain, how you doubt it, and how you connect it to your own judgment and action.

In this article, I want to think through how we should gather information in the age of AI, how we should question it, and how we should put it to use in our decisions.

The “speed first” information environment shaped by AI, social media, and the press

AI makes information gathering far more efficient. It organizes multiple angles, summarizes the key points, and lays them out in a form that is easy to compare. There are more and more moments when you can grasp the whole picture faster than you could by researching alone.

At the same time, an answer produced by AI is not a verified, guaranteed fact. An AI response can contain errors in the underlying source, out-of-date material, missing context, or shifts in interpretation. Precisely because it is convenient, treating that output as settled information is not a responsible judgment.

Social media is also powerful in terms of sheer speed. You can quickly pick up voices close to the front line and the reaction of the wider public: what is drawing attention, what dissatisfactions and expectations are surfacing, and what the general mood feels like. For sensing these things, it is useful.

But the loud voices that stand out on social media do not necessarily represent the whole reality. The loudest people are often the most visible. Opinions that are easy to inflame tend to spread. Skewed views sometimes gather the most impressions.

Even established media such as newspapers and television are affected by an environment that demands speed. The more immediacy is required, the more the first report can go out before the information is fully organized. In accidents and disasters that draw attention, it is not unusual for the first report and later reports to differ.

The point is not that AI is bad, or social media is bad, or the press is bad. When you come into contact with information, you need to consider not only where it came from, but both how accurate it is and how you use it.

Treat fast information as “not yet complete”

It is not that early-stage information has no value. On the contrary, there are moments in the first response where fast information is important. Even in incident response, it is the first report that lets you estimate the range of impact and set the priorities for investigation.

Suppose an incident occurs in some service. In the earliest stage, both the cause and the range of impact are often unconfirmed. In that situation, the following can be material for a rough first read: inquiries from users, monitoring alerts, reactions on social media, and internal reports.

Still, a first report is only a first report. The information may not yet be complete, additional details may emerge later, and the content may be revised. Treating a first report as the basis for a final judgment without verifying it is not a responsible decision.

And yet, unconfirmed information still has its uses: grasping a trend, forming an initial hypothesis, estimating the range of impact, listing the angles to investigate, and organizing what needs further confirmation. For these purposes, speed matters more than accuracy.

The problem is not using early-stage information. The problem is treating information that is not yet complete as if it were confirmed.

A breaking report, a reaction on social media, or an AI summary can serve as an entry point to a judgment. But turning it into the conclusion of that judgment requires verification.

Read information as “fact,” “opinion,” and “speculation”

When handling information, it is important to separate fact, opinion, and speculation. When these three are mixed together, judgment goes wrong more easily.

Say there is a report that “the service is hard to connect to.” If monitoring actually shows the error rate rising, that is a fact. “Maybe the cause is the database” is speculation. “The operations response was poor” is an opinion.

If you handle these without separating them, you can end up treating a loud opinion as if it were a fact, or advancing your response on the premise of a cause that is still only speculation. It is not enough to gather information; you need to sort it by type.

Using AI to classify the information you have collected into fact, speculation, and opinion is effective. But that classification is itself an AI output. In the end, a human has to check the original information and review whether the classification is sound.

The value of information is not decided by accuracy alone

Let me pause here and consider what “accuracy” of information really means.

When we talk about accuracy, we tend to mean how close something is to the facts and how few errors it contains. That matters, of course. If you judge based on information that has not been adequately fact-checked, the judgment itself goes off course.

However, it is not the case that only high-accuracy information has value.

To begin with, much of the information we receive is a mix of fact and opinion. This does not necessarily mean it is misinformation. It happens because the perspective of the person who saw it, conveyed it, and organized it enters into it.

Even for the same incident, what people see changes with their position:

• Engineer: the cause and the logs
• Customer: the range of impact and the estimated time to recovery
• Manager: accountability and recurrence prevention
• Sales: the effect on the customer relationship

Each of them is looking at something different. That is exactly why different information emerges about the same event.

So does value lie only in bare facts with no filter of perspective applied? No. By seeing an event from many perspectives, the event becomes three-dimensional. Impacts, background, and how those involved received it, things that a simple list of facts would not reveal, start to come into view. That is where depth of information is born.

Searching for information that looks useful, and trying to raise the accuracy of information: neither of these is wrong. In fact, they are basics of gathering information. But it has become hard to create difference or value from those alone.

With AI, anyone can now organize information to a reasonably high standard. With a search, many people reach the same primary information. Look at social media, and you can pick up similar reactions and opinions. In other words, the scarcity of gathering information itself has fallen.

So where does the difference come from? It comes from how you think about the information you encounter, how you translate it into your own work, and what you confirm next and which judgment you connect it to.

The value of information is not decided by the information alone. It changes with how the person who received it uses it. That is why, from the information-gathering stage onward, thinking about “how to handle information” matters even more than “collecting only correct information.”

Real problems are made of multiple factors

When a problem occurs, we want to narrow the cause down to one thing: what went wrong, who made the mistake, where to fix it. If you can reduce it to a single cause, it looks clear and easy to deal with.

But real problems are not that simple. Even a system malfunction does not always have a single cause. And in business or organizational problems, multiple factors are tangled together in complex ways.

For example, even when considering why inquiries increased, the product malfunction alone is not necessarily the cause. Insufficient notice of a spec change, a hard-to-follow manual, a change in the customer’s own operations, inadequate support capacity, a gap with what was explained during sales: several factors may overlap to produce the result of rising inquiries.

That is why information gathering is not only for finding “the single cause.” It is for surfacing multiple factors and gauging how much each factor seems to be contributing.

In real problem solving, you do not derive a single correct answer. You confirm the factors one by one, apply measures, and watch the results, working toward a resolution.

High-accuracy information is close to the primary source

AI, social media, and news sites make information gathering more efficient. But that information has, in most cases, passed through something someone saw, heard, or interpreted. The further you get from the source, the more room there is for interpretation and error to slip in.

To raise accuracy, you need to move closer to the primary source: official announcements, contracts and specifications, system logs, monitoring data, the actual screen, direct confirmation with the customer, hearings with the people on the front line, and the results of trying it yourself.

It may look like the long way around, but in the end it is easier to raise the quality of your judgment this way.

Precisely because digitalization has advanced, the value of finally confirming things with your own eyes and feet has grown: looking at the screen, seeing the site, confirming directly with the people involved, and trying it yourself. These may look like analog methods, but in the sense of getting close to the information you need right now, they can be the highest-accuracy methods available.

AI should be used not as a substitute for the primary source, but as an aid for getting closer to it. Use it as the entry point to the work of confirmation, such as what to check, which materials to look at, and from what angle to read the logs, and AI becomes a powerful tool.

Design decision-making and risk response on the premise that information can be wrong

No matter how carefully you gather a large amount of information, accuracy almost never reaches 100 percent. Even as you move closer to the primary source, oversights and misreadings can happen. That is why information has to be treated as something that can be wrong.

When making a decision, you need to consider not only the information itself, but also: how accurate that information is, where errors are likely to slip in, who would be affected and how if it turns out to be wrong, at what point you will re-confirm it, and how you will respond when you notice an error.

In a system release, you consider whether you can roll back if a problem occurs. Judgment is the same. If the information turns out to be wrong, how far back can you go, and from what point can you no longer return? Proceed without considering this, and a failed judgment can lead to a large loss.

Of course, in reality there are also judgments that cannot be rolled back as easily as a system: explanations to customers, committed budget, staffing assignments, changes in business direction. Once these are set in motion, they cannot be fully reversed. That is exactly why you need to think in advance about the impact of misjudging the accuracy of your information.

What matters is not eliminating failure entirely. It is factoring in the possibility that information is wrong, and identifying the range in which things are still recoverable. You need to build in margin and buffer, and create a state where even a failure can be turned into a lesson for next time.

Even if a judgment was wrong, if you are in a state where you can learn from that failure, the accuracy of your next judgment improves. So that you can say “that failure is why I am where I am now,” you need a design in which failure does not become fatal. Information gathering, decision-making, and risk response should all be considered together, to that extent.

Conclusion: what is questioned is the ability to use information while doubting it

Thanks to AI and social media, information has become available quickly and in large volume. That in itself is a great value. As the speed of information gathering rises, the first response gets faster and the options increase.

But the earlier the stage of the information, the more you need to estimate its accuracy and how likely it is to be updated. A breaking report, a reaction on social media, or an AI summary can be an entry point to a judgment. But there is danger in making it the basis for a final judgment as is. When handling information, you need to separate whether it is fact, speculation, or opinion.

Who is saying it. How far it has been confirmed. What level of accuracy it has. Judge without looking at these, and you get swept along by the speed of information, and the quality of your judgment declines.

That said, gathering high-accuracy information is not the only value of information gathering. In an era where many people can gather information in the same way, it is hard to create difference simply by possessing information. The difference comes from how you read it, how you doubt it, and how you connect it to your own judgment.

To raise accuracy, you need to move closer to the primary source. And because information can still be wrong, you also need to consider how you will respond if it is. Judgment includes not only information gathering, but also the risk design for when the information turns out to be off.

What is questioned in information gathering in the age of AI is not the ability to believe information. It is the ability to use information for judgment while doubting it.

Do not get swept along by speed. Estimate accuracy. Decide on the premise that things can be wrong. And consider how to translate the information you obtain into your own work and judgment.

That posture, I believe, is what will support judgment in the age of AI.