With the help of the Internet、The development of artificial intelligence and computer science，Calculating the rise of social sciences does bring a lot of breakthroughs and expansion to social research。Use new methods such as online media and cloud computing to obtain and analyze data，Especially the "computer algorithm" that enables the data and empowerment through "machine learning"，It provides an unprecedented new paradigm or thinking method for research and explanation society and predicting society。This is also the main reason for the development of social science as an emerging field of academic sciences so rapidly。But，Research on a complex society，Calculation is still difficult to avoid some limitations，Understanding these limitations helps the theoretical and methods of using the calculation to develop social research。

First，Calculating data difficult to cover the relationship between the constituent elements and elements of the complex society。Data in traditional social surveys are difficult to cover all kinds of people in society，For example, a sociologist Eric Olympic Wright specifically pointed out in the preface of his book "Class in the Poor Industrial Society"，"The irony is ironic is，Inside Marxist tradition，Critical criticism of capitalism is to directly aim to the richest part of the capitalist class，The moral condemnation of capitalism is largely based on it that makes poverty permanent existence Bet365 app download "。And sampling survey is due to limited conditions，Two extremes in the class structure have been missed: the real owner of the capital (big capitalist) and the "underlying class" on the edge (the unemployed and the person who lost labor)。This kind of omitted important sample situation，Scholars who believe in data survey will not be unfamiliar。

There is also such a problem with the method of using the new data，For example，The poor who buy a car and buy a house need to be borrowed，But there is no record in the bank，If the credit provided according to the algorithm is implemented, borrowing，Then these people will not get a loan; for not using Facebook、Weibo、Twitter、WeChat and other social media personnel，When calculating and analyzing the data recorded by these social media records，It will also be missed; although the Google Book Library already has a large amount of data for hundreds of years of books in many countries in the world，but it is just a part of it，How much is the representative，The analysis results are still difficult to solve the problem。According to the "Digital 2021: Global Overview Report)，As of January 2021，5.22 billion people in the world use mobile phones，equivalent to 66.6%of the world's total population，It also means that one -third of the population does not use mobile phones。The number of people in the world of global use reached 4.66 billion，less mobile phone users，Social media users account for half of the global total population。Another，Some social characteristics are important，But it is difficult to quantify，It is also difficult to enter the calculation range。Information collected with the Internet as the bet365 best casino games main channel，Although the sample volume is large，But the representativeness of the coverage is more difficult to control than the traditional social survey method。

2，The calculated algorithm is the logic of processing information，Even if the data is available，What algorithm is still a big challenge。Traditional quantitative research methods focus on explaining the relationship between an independent variable and the cause variable，Especially identifying its cause and effect。Use data training to obtain a model for prediction，It is the big data algorithm。This algorithm for forecasting will use the specific settings of the model as a black box，As long as the prediction of the forecast variable is more accurate, the better。For example, the problem of gender discrimination of high -tech industries，The report given by Google in 2017 states that female employees are obviously lower than male income，But after controlling the promotion and subjective performance evaluation variables，The income of Google employees has no gender differences。The increase or decrease of the control variable is also the source of the model uncertainty。If the algorithm that does not increase these two variables is adopted，The income gender differences of Google employees have appeared，The difference after the increase will disappear。Gender discrimination lies in promotion and subjective performance evaluation，Under the same conditions, men are more likely to obtain subjective performance evaluation of promotion and good performance。If the cause and effect analysis of gender discrimination is to be done，The purpose of the algorithm is not just the best fit model to predict the income of employees。How should bet365 Play online games the reporter choose the algorithm? Is the variables such as a strong revenue such as evaluation and promotion?。

The uncertainty of the model provides some researchers with the opportunity to "pick cherry"。Assume that there are 13 control variables in a model，Different combinations with independent variables，will produce 8192 different models，This is 8192 different algorithms。Suppose we want to evaluate the effect after the implementation of a policy，If there are a dozen control variables，In thousands of algorithms，The independent variable coefficient may have two basic situations: (1) All are positive (or negative value) and statistically significant (or partial significant); (2) some estimation coefficients are positive values，Some are negative value，Some remarkable，Some are not significant。The assumption coefficients are all positive and significant，So which report should the reporter choose? What if there is a positive value?，The distribution of all coefficients is published，Put the results of the report model in the distribution chart，Present it to the reader。Calculated logic requires people to explain and give meaning，Numbers and algorithms itself does not make sense。Except the model selection question，The uncertainty of the algorithm also exists in data collection、Cleaning and indicator construction, etc.。Claudia Wagner and others published a paper "Society of Measuring Incorporating Algorithm" in the "Nature" magazine，Discussion When individual push this type of algorithm spreads according to the characteristics，The impact on the measurement of the measuring (error) measurement of society，Disposal quality lack Bet365 app download of measurement。So，Computing Social Sciences also faces the challenges of how to ensure the effectiveness and trust of measurement。

third，The uncertainty of the emergence of the dynamic evolution of algorithms and the dynamic evolution of the complex society and the factors behind the predictive algorithm have brought the limitations of calculation in complex social research。The original intention of the artificial intelligence pioneer Herbert Simon to study artificial intelligence is to solve the problem of people's decision -making。Winning the Nobel Prize in the field of economics is derived from the limited rational and satisfactory theory he proposed，Change the origin of the economics that the original rationality of people and seeking benefits to maximize the hypothesis。

Under a complex social environment，The challenge we face is invisible ahead，I don't know where you are; there are dangerous factors; the environment is not static，but dynamic。Under such complex social conditions，There will be emerging phenomena，Any subtle changes may cause unpredictable large new phenomena，It is difficult to identify the cause and effect。The emergence of emergence cannot be simplified to the underlying law，It is difficult or even predictable。The complex "system" involves not only game rules (algorithms)，There are also actors and their decisions made on a large number of options on each choice point。In a complex society，The limited result of people's limited rationality and data and algorithms is，Even if there is real -time data, it is difficult to accurately predict the next step (not to know whether the optimal)。We know that the "greedy algorithm" that looks bet365 Play online games best step by step is not desirable，Because in the end, it is often not the best solution of the overall situation。Sometimes，Life is like a function with many extreme points，Stand at a maximum value point with zero guide，Although it looks like taking it down, it looks like going down，But it is often the seemingly irrational、take a step down，A peak that allows you to discover a higher mountain。Modern society with such high complexity，The seemingly rational "greedy algorithm" is even more unavailable。In terms of model settings，Professor Christopher Win Shipu, Harvard University, think，Intuition、Theory and facts and luck are helping us discover the best model in a certain sense。That is to say，Although we have a very advanced computing technology，Continuous development of new calculation software，But the luck component contained in the decisive factor of the algorithm itself is difficult to completely control。The accidental factors in the complex society exist at all times of luck ingredients，The setting of algorithms in social research has caused uncertainty that is difficult to solve。Even the model -based modeling，ABM (Agent-Based Modeling)，As a simulation technology proposed to solve complexity，It can only be limited to the rules of the main interactive game。

The algorithm may only help us find certain certainty for uncertainty，Behind it is human emotion and preference。Each decision in life is used as data，Train "Model"，Cognitive mode，So make decisions every day and take action。The algorithm is in our daily life。As a computing social sciences，We must understand more than ordinary people and calculate the bet365 live casino games limitations of research in the face of complex society，Innovative services for social science methods and theoreticals can be better used，Then make due contributions to building a beautiful society。

　　bet365 Play online games