Big data, it is not allowed to be the old man

#This article is produced by the product manager's "Original Incentive Plan". After entering the summer, the weather began to become "impermanent", and the storms were stunned and the sky was clear for a while. Before going out, look at the weather in your hand and find the same mobile phone model, the same weather software, and the same time and place, and there are different weather prediction results. Why is this? The author of this article analyzes this, let's take a look.

1. The same eaves, different weather

Summer is here, and it is rainy.

It's okay before work. A group of boring people looked at the sky outside the window and bet on a few points to decide who paid for lunch.

While betting, someone had picked up the phone and started to check the strategy.

In fact, this is not cheating. Everyone knows that the weather forecast software is not allowed to report the weather, and sometimes it is not even accurate.

Moreover, colleagues hold different mobile phones and use different weather forecast software. The results of the weather predictions are really different.

But when my colleague Xiao Ai, when I read the prediction results of her mobile phone, some of my expectations -Xiao Ai and I used the same mobile phone. The weather software is the weather software that comes with the system, but the weather forecast is actually with me. Different from the phone.

I brought her mobile phone and confirmed again: We are in the same position, we are setting up systematic positioning, and we all give the weather software to read the positioning authority. We are also in the same WIFI environment.

We use the same mobile phone and the same weather software to update the same version.

In order to avoid time errors, we have refreshed the weather forecast page many times.

But the weather forecast results are still different.

At least one of these two weather forecasts is wrong. Or both of them are wrong.

Does the data push of weather forecast also give different results according to the love of different people?

Second, the same trace, different positions

In recent years, many colleagues have chosen bicycles to get off work.

Because of health, because of environmental protection ... it may also be because of poverty.

There are several important offices near the company. The management is relatively strict. The shared bicycles cannot be parked anywhere. They must be placed in the designated area.

In specific operations, it must be operated on the mobile phone now, and confirmed that it has arrived at the regulations to return the car to the lock.

Then, those colleagues who love to sleep laziness have experienced what nightmares are.

There are still 2 minutes before the work time. Riding a bicycle to the company downstairs, stop in the parking area, click "I want to pay back the car" on the phone, and then see the prompt: you are not in the car returning area.

Later, naturally, it was a toss, moved forward, and moved backwards, but no matter how to enter and exit, the mobile phone page always displayed: 3 meters away from the specified return area.

What's more, the people who came to the parking next to each other arrived at the position. Many of them directly closed the locks and did not encounter trouble in the positioning link.

Seeing that the time of work has passed, the collapse of adults is often in that moment: our positioning data, why will it always be 3 meters?

However, he was also relieved quickly: Since the car could not stop, just ride a bike out for breakfast. When you are late, you have to deduct money, and you have to deduct money in violation of regulations. Anyway, you always have to deduct one.

Third, the same search, different results

Still living in the office.

At 5 pm, people were busy finished the task of hand, began to swipe their mobile phones, and waited for get off work.

Xiao Aixin bought a headdress, and Yang Mei looked good and asked her to link.

WeChat forwarding Taobao link is more troublesome. Xiao Ai is a lazy person, so he told Yang Mei: You search for#$ %% @@%. The first one is.

Yang Meizi obediently made a obedient photo, and entered#$ %% @@%in the Taobao search bar, but not to mention the first one, I brushed ten pages, and I didn't see the headdress I bought.

In fact, different people search for the same word on Taobao will have different results, and it is not new.

The system will recommend them based on everyone's search and purchase habits. The system believes that they are most likely to be interested and most likely to buy.

However, these recommendations have obviously not completely guess the user's intentions.

The search results are not interested in Yang Mei.

The goods she wanted to buy now, but the system did not push her.

Is the judgment of big data very inaccurate?

Fourth, data defects, obvious

Is there a problem with big data?

It is far from perfect.

In these trivial things in life, we can also see that big data is obvious flaws.

1. Because of the edge data, the main problem is ignored

Today's big data is calculated based on a huge amount of data.

However, there is a difference between data and data: some data have an important impact on the calculation results, but some data are only used as a reference.

But when marginal data is enough, it will also have a qualitative impact on the results. Let some data that should have played a decisive role is marginalized.

Just like the difference between the weather forecast results just now, it may also be the difference between data calculation.

Not only are the forecast of the Meteorological Bureau, but also combines individual differences in individual differences in the action trajectory and rain frequency of each person.

But what is the analysis result?

At the same time, at the same place, it gave a completely different weather forecast.

Is it raining, will it change according to the trajectory of everyone?

We do not live in the world of Chumen, and we are not called Xiao Jingteng. Under the same roof, the same weather should be faced. This common sense is squeezed aside by a large amount of data in the calculation.

In the online world, the facts are defeated by a lot of data, which is actually very common:

Praise a lot of videos may not be really interesting.

If you can find enough people to brush, pay attention, praise, and comments, some low -quality videos can also become popular and push to many people.

The restaurants with low scores may not be difficult to eat.

Just find enough people to go to malicious bad reviews, you can easily get a restaurant's score. At the beginning of the opening of many restaurants, they will also find a lot of people to praise them and let themselves stand on the recommendation head.

When there is enough true evaluation, the truth will be covered. As for the taste of the meal, the data cannot be tasted.

2. In a complex way, make simple questions

Because of the existence of big data, many simple problems that were very simple became complicated.

Some obvious answers were hidden.

When you search for 1+1, 80%of the search results on the first screen are not 1+1 = 2, but the Gothic Bach guesses, movies, songs, etc. are related to 1+1, and even about 1+1 = 1 inference.

Obviously a very simple problem, but because of the existence of a large amount of relevant information, it has evolved very complicated.

Just like Xiao Ai recommended by Yang Mei, the two of them have used the same WiFi address for a long time, and they also pay attention to each other and add friends with more than one software. One of them purchased a certain product a few days ago, and the other used the same search words to search the same product a few days later. Presumably, the data could be monitored.

If you only consider these obvious information and recommend the same product, you can immediately promote the purchase.

However, big data considered too many past search and purchase results. They thought that they were not the same consumer group and thought that they had different preferences, so the search results for pushing were completely different.

A simple question, too complicated.

3. Data bias, bias spiral

Big data push has more serious prejudice.

Data knows users, often like blind people touching.

Only when I touched the elephant's legs, I thought the elephant was a cylindrical body.

The data only sees the user's side, thinking that the user is like that.

Initially, the data recommended some products and videos to Yang Mei through blind guessing. She watched a few of them at will.

The data will record her watching behavior and push similar products and videos again for verification.

Sure enough, Yang Meizi was really interested in these contents and watched many times.

After a long time, the data will remember: Yang Mei is a young girl who likes hot pot, likes young fresh meat, and likes punk style. And will push her more related content.

Although this kind of push is right, it is just a prejudice.

In fact, Sister Yang also likes Chinese style painting, green headdress, and northeast cuisine, but the data does not know.

The data is not omnipotent and omnipotent, and there is no way to know the information you have not collected.

Especially in targeted push, a prejudice spiral will be formed.

The system kept pushing Yang Meizi hot pot, punk, and small fresh meat. Yang Mei did like to watch these, and continued to watch, like, comment, and purchase.

Data records in the case will continue to increase related recommendations.

Finally form a cycle:

But for a person, what you like, you have seen it for too long, and occasionally you are tired. The system is still happy to push these contents.

Because of enough data to confirm, she was interested. Even if I did not watch it once or twice, in the overall data, these two behaviors that did not watch or not still could not resist the frequent preferences before.

Unless she is really hysterical, I ordered dozens of times "don't recommend similar content". But most people do not do that. After all, the recommended things are also interested in themselves, and they are worried that they will not be seen in the future.

In this prejudice cycle, the more content that has determined the preferences, the less the other content can meet with the user. Therefore, the opportunity to complete the prejudice of data becomes less.

While pushing hot pot, punk, and small fresh meat to Yang Mei, a popular video was also pushed to her eyes, which was painted about Chinese style.

But the protagonist of this video, talking about the mother, is just an unacceptable type that Yang Mei is unacceptable. The data happened to be lost once, and she learned that she liked Chinese style painting.

I will have such an opportunity next time. I don't know if it is a few months and a few years later.

Five, big data, not big enough

All these issues are in the final analysis that big data is not big enough.

Limited by the reasons for technology, cost, and privacy, there are not enough ways to obtain data and not accurate.

If the amount of data capture is large enough, it is not limited to a platform, but not only limited to the virtual world, and can see more details. Data has higher opportunities to really know a user instead of generating prejudice spirals.

If the data acquisition method is detailed enough, all the data acquisition devices are accurate enough, and there will be no problems that cannot be repaid.

If the detection method of big data is high enough, the unpredictable weather will one day be able to judge that it is not bad.

But due to technology, big data cannot be seen at a glance at a glance. Only through the way of peeping in the leopard in the tube, the field of view of the field of vision observed by hundreds of millions of angles can be made up for the portrait of a leopard. Limited by cost, big data cannot obtain sufficient data. Perhaps hundreds of millions of vision fragments can spell a pantra of a leopard, but the cost only allows tens of thousands of fragments, and the leopard that is spelling is naturally speculative.

Limited by privacy, big data cannot obtain some key data. Among the tens of thousands of fragments, there are also fragments of some key parts of the leopard.

The end result is that the panther portrait made of big data, although the prototype of the leopard, is far from the real leopard.

Six, contradictory privacy

The limitations of technology and cost, with the development of the times, there is always a day of resolution.

Just like the traditional concept of Tianyuan, one day will be subverted under the witness of astronomical satellites.

However, the future development of big data will inevitably have more intense collisions with our privacy needs.

Especially our real needs, even it is difficult to say that most people's attitudes towards big data, want to refuse to welcome.

When big data prediction is needed, we require big data to maintain accurate.

When we do not need big data prediction, we are worried about leakage of privacy.

Take a taxi to a remote corner, and even we don't even know how to search for the target location, the big data directly predicts our target location. This saves our time and makes our travel simple and easy. But after getting in the car, we began to worry again. Will our travel information be used by someone? The products we want to buy are directly pushed to the e -commerce platform to save our search, browsing, and comparison time and energy. But while enjoying this convenience, some people are worried about whether they will be known to others. We are all contradictory. We want to possess good data, but we want to abandon everything that is not good.

But the good and bad of big data is the opposition and unity.

Without enough data to bury points, more accurate predictions cannot be made.

Without behavioral monitoring again and again, you cannot send the information you need to be in front of you when you need it.

To put it easier: Navigation software in mobile phones, without accurate positioning, no one knows where you are, where you go, and how can you give you accurate navigation?

As some people say: "In fact, we are not without the right to choose. Even in modern times, we can throw away their mobile phones. One person goes to the mountains to live in the mountains to protect our complete privacy."

Although this is a bit cool, it can also reflect a problem: When we replace non -smartphones with smartphones, when we enjoy the convenience of remote shopping, when we searches for answering our doubts at any time, we can search for our doubts at any time, and We should also know that we will come up with a part of privacy to exchange such convenience.

Convenient privacy exchange is the inevitable result.

In the process of big data development and perfect, what we need to fight for, society needs to be regulated, is how much privacy and how much it can be exchanged; the public pays for privacy. Where is the border; in addition to the convenient exchange of privacy, will it be seen and used by others.

There is no answer yet, but sooner or later there will be answers.

One day, big data can be considered the weather.

One day, we can find convenient and privacy balance.

#Columnist#

Mo Yan, everyone is a columnist, online marketer, and psychological counselor. Good at consumer behavior, text communication, marketing and other fields.

The original published in this article is a product manager, and reprinting is prohibited without permission.

This article is produced by the product manager's "Original Incentive Plan".

The title map is from Unsplash, based on the CC0 protocol

- END -