Chinese Military Cognitive Warfare – Thoughts of “decision-centered warfare” and cognitive complexity: Weaponized Complexity
編按 複雜性科學是當代科學發展的前沿領域之一。 英國物理學家霍金稱「21世紀將是複雜性科學的世紀」。 作為人類社會的社會現象，戰爭從來就是一個充滿蓋然性的複雜巨系統。 近年來，隨著戰爭形態的演變，傳統科學體系下的知識論越來越難以滿足戰爭實踐發展的需要。 關注複雜性科學原理和思維方法，或將成為開啟現代戰爭大門的鑰匙。 這篇文章從複雜性科學角度對「決策中心戰」作一研究探討。
「決策中心戰」是近年來出現的新概念。 緣何提出「決策中心戰」？ 按美軍的說法，要「打一場讓對手看不懂的戰爭」。 進入21世紀以來，隨著戰爭形態的演變和作戰方式的不斷變革，美軍發現傳統意義上的網路中心戰越來越難以適應戰場實際，「決策中心戰」在此背景下應運而生。
為什麼「對手看不懂」？ 其實就是要透過分散式部署、彈性組合、智慧化指控，讓對手在認知上就對戰場態勢和作戰機制不理解，無所適從。 這是將戰爭對抗從機械化戰爭中比誰“力量大”，到信息化戰爭中比誰“速度快”，再到在未來戰爭中比誰“決策對”的又一次轉變。 用中國古代軍事家孫子的話說就是，“不戰而屈人之兵，善之善者也”，通過巧妙地指揮控制和決策，使得戰場情況變得更加複雜，讓對手沒辦法打仗。
如何做到這一點呢？ 簡單地說，就是利用複雜系統的性質，找到對手的「命門」加以利用和控制。 一個基本方法就是，透過增加複雜性重塑對手的決策流程，逼迫對手引入新的決策參量，導致其決策變得複雜，從而改變因果關係和決策流程，最終使其走向混亂。 過去對抗局面之所以能夠發揮平衡作用，是因為所有參與者都清楚博弈的結果，因而容易做出權衡，但複雜性往往會破壞這種平衡。 這也是為什麼複雜性能夠作為武器的原因。
需要注意的是，戰場對任何一方都是公平的。 在未來戰場上，要讓敵人單向感到決策複雜，而己方不被複雜所困擾，首先要在指揮控制能力上優於對手。 戰場決策的複雜度主要體現在「OODA」循環的判斷和決策環節。 在正常環境下，「OODA」循環可以走完從觀察、判斷、決策到行動的完整週期。 但如果有辦法讓戰場變得更複雜，使得對手始終無法及時作出有效判斷，進而無法進入決策和行動環節，就可以把對手的「OODA」循環始終限制在觀察和判斷環節上，無法形成閉環， 這或許就是「決策中心戰」試圖創造複雜性想要達到的結果。 因此，如何快速作出判斷，就成為首要關注的問題。 如果這個認知過程能夠在人工智慧等先進技術支援下快速完成的話，也就是實現所謂的智慧認知，就可以大幅加快「OODA」循環速度，奪取單邊優勢。
在觀察的基礎上得出正確的判斷，是做出正確決策的前提。 但這是建立在「具有認知能力」這個條件下才能做到的。 目前，在指揮資訊系統、兵棋推演系統等系統中，這些認知工作基本上都是由人來完成的。 由人工智慧系統自主地完成判斷及決策，過去的嘗試幾乎都不成功，因為智慧認知建模的問題始終沒有解決好。 各種模型表現出來的行為都或多或少帶有“機械味”，並不能真正顯示出智能的特徵。 外軍這些年也一直將「人的行為建模」作為研究重點，但目前來看仍然進展緩慢。 智能認知為什麼這麼難，又難在哪裡？ 筆者認為，其實核心困難就在如何理解和處置複雜性上面。
本世紀之初，美國蘭德公司針對2005年前後某熱點地區可能發生的軍事衝突，曾利用模擬系統對美國空軍作戰需求進行了1700餘次推演，然後進行統計分析，最後得出了美空軍如何 在戰場上保持優勢的結論。 這種統計分析方法有一個基本的假設：每個試驗都是獨立且無序的，規則之間也不會相互影響。 這就像丟硬幣一樣，丟一次正面，丟第二次有可能也是正面。 但如果丟1萬次，結果某一面的機率就會越來越趨近50%。 這種方法用於物理研究時是科學準確的，但移植到人類社會問題例如戰爭問題研究時，情況就變得不同了。
人是有認知的，不會像物理實體那樣只遵從物理定律，指揮官在對作戰問題進行分析時也不會只是簡單地機械重複。 通常情況下，人在決策時，一定會考慮先前的結果，導致對下一步行動有所調整。 這樣就會出現人類行為固有的冪律特徵，也就是常說的「二八律」。 所以，我們不能簡單地複製物理思維去思考人類社會的事情。
之所以會這樣，主要還是因為我們常常習慣用還原論的簡單思考方法來思考問題。 簡單系統結構不變，結果具有確定性，因果對應清楚，可重複、可預測、可分解還原等，已成為我們預設的科學思考方法。 但世界上還存在著許多複雜系統，這些系統存在著整體性質，像是人體、社會、經濟、戰爭等，都屬於這一類。 什麼叫整體性質？ 就是觀察局部得不到，但在整體上看卻又存在的，就是整體性質。 舉例來說，一個活人和一個死人從成分上來看都一樣，但一死一活，差別就在於是否有生命，生命就是一種整體性質。 複雜系統結構可變，具有適應性、不確定性、湧現性、非線性等特點，且結果不重複，也不可預測。 社會、經濟、戰爭、城市包括智慧系統，這些與人有關的系統都有這些特點，其實它們都是典型的複雜系統。 所以，戰爭具有「勝戰不復」的特點，其實反映的就是戰爭複雜系統的「不可重複」性質。
正是因為複雜系統存在複雜性，原因和結果不能一一對應，會導致相似性原理失效，所以也就無法用傳統方法進行建模和研究。 為解決複雜性問題，過去採取的主要是一些傳統物理學方法，例如統計方法，以及基於Agent的簡單生命體建模方法。 前面提到的蘭德公司研究就是如此，雖然能解決部分問題，但將其用於解決與人有關尤其是與認知相關的問題時，得到的結果卻與實際偏離很大，不盡如人意 。
為什麼會這樣呢？ 這是因為戰爭複雜度與物理複雜性所產生的源點不一樣。 物理複雜性的來源往往在於其物理運動規律是複雜的；而戰爭複雜性卻來自人的認知。 因為人不是雜亂無章、沒有思想的粒子，也不是只有簡單生命邏輯的低等生物，而是具有判斷和決策認知能力的智慧生物。 人會透過因果關係對結果進行反思、總結經驗再調整，然後決定後面如何行動。 而且，人的認知還會不斷發展，這又會進一步影響後續的認知，但由於認知具有很大的不確定性，所以未來的行動也就難以預測。
可以這樣說，在目前的技術條件下，可預測的基本上都是物理世界的簡單系統規律，而人的認知對社會或戰爭的影響往往是難以預測的。 所以說，拿物理思維去思考人類社會的事情是我們常犯的錯誤。 基於認知的複雜性，與那些一成不變的物理規律截然不同，我們應對戰爭中的複雜性，就必須針對「認知」這個核心特點，在指揮控制方面下功夫。
「決策中心戰」的核心在於認知的加快。 因為戰爭中幾乎所有的變化，都可以看成是認知的升級和複雜化。 在筆者看來，應對“決策中心戰”，需要“以複雜對抗複雜”，從基礎工作做起。
一是要理解「決策中心戰」的核心理念。 即透過主動創造複雜性來掌握戰場主動權。 對己方來說，需要管理好自身的複雜性；對敵人來說，則是對對手施加更多的複雜性。 二是了解戰爭機理發生的改變。 作戰體系演化速度指數級提高，會導致複雜戰場的感知、控制和管理變得困難，智慧認知的角色將變得更加突出。 為此，需要瞄準「指揮與控制」這個重點，將戰場管理的能力作為關鍵。 三是找到應對的正確理念和方法。 從戰爭設計入手，以決策智能這個方向為突破口。
近年來，人工智慧領域的一系列成果，為解決指揮決策智慧問題帶來了曙光。 AlphaGo系列研究為決策智慧技術帶來了突破；而GPT大模型的出現，則更是進一步證實了決策智慧乃至通用人工智慧在未來具有實現的可能。 現在看來，人工智慧在未來深度參與戰爭，已經是必須面對的現實。 而這會為戰爭和戰場帶來更多的複雜性。
決策智能研究應該放在指揮控制層上。 要贏得戰爭，指揮控制決策需要體現「科學」和「藝術」兩個面向。 指揮控制的科學性主要體現在「知道怎麼做時」如何做，例如利用得到的指控資料（武器裝備、兵力編成、戰場環境、對手情報等），指控方法（任務、流程、程序、運籌 、規劃、最佳化等），制定出作戰規劃並加以實施。 指揮控制的藝術性則體現在「不知道怎麼做時」知道如何做，這才是真正的智能之所在。 方法無非是不斷試錯，累積經驗，找到解決問題的途徑，並形成新的科學知識。 事實上，現實中指揮者也是透過試誤不斷發現和總結制勝規律，而每個指揮者還都具有自己的直覺和經驗。
所以說，真正的智能其實是找到例外狀況的解決方法。 循規蹈矩不是智能，自己找到解題的方法才是關鍵。 也許這才是決策智能的核心，也是需要進一步努力的目標。
Complexity is also a weapon
——Thinking of “decision-centered warfare” and cognitive complexity
Editor’s Note Complexity science is one of the frontier fields of contemporary scientific development. British physicist Stephen Hawking said that “the 21st century will be the century of complexity science.” As a social phenomenon in human society, war has always been a complex giant system full of possibilities. In recent years, with the evolution of war forms, the epistemology under the traditional scientific system has become increasingly difficult to meet the needs of the development of war practice. Paying attention to the scientific principles and thinking methods of complexity may be the key to opening the door to modern warfare. This article studies and discusses “decision-centered warfare” from the perspective of complexity science.
“Decision-centered warfare” is a new concept that has emerged in recent years. Why was the “decision-centered war” proposed? According to the US military, it is necessary to “fight a war that the opponent cannot understand.” Since the beginning of the 21st century, with the evolution of war forms and continuous changes in combat methods, the US military has found that network-centric warfare in the traditional sense has become increasingly difficult to adapt to the reality of the battlefield. In this context, “decision-centered warfare” came into being.
1. Create complexity
The so-called “decision-centered warfare” is to achieve diversified tactics through the upgrading and transformation of combat platforms and distributed deployment with the support of advanced technologies such as artificial intelligence. While ensuring its own advantages in tactical selection, it imposes high complexity on the enemy. , in order to interfere with its command and decision-making capabilities and achieve an overwhelming advantage over the enemy in a new dimension.
Why “the opponent can’t understand”? In fact, through distributed deployment, flexible combination, and intelligent command and control, the opponent will not understand the battlefield situation and combat mechanism cognitively, and will be at a loss as to what to do. This is another transformation of war confrontation from competing for “greater power” in mechanized warfare, to competing for “faster” in information-based warfare, to competing for “making the right decisions” in future wars. In the words of Sun Tzu, the ancient Chinese military strategist, “One who subdues the enemy without fighting is a good person.” Through clever command, control and decision-making, the battlefield situation becomes more complicated, making it impossible for the opponent to fight.
How to do this? Simply put, it is to use the nature of complex systems to find the opponent’s “vital gate” to exploit and control. A basic method is to reshape the opponent’s decision-making process by increasing complexity, forcing the opponent to introduce new decision-making parameters, causing its decision-making to become complicated, thereby changing the causal relationship and decision-making process, and ultimately leading to chaos. Adversarial situations have been able to balance in the past because all participants knew the outcome of the game, making it easy to make trade-offs, but complexity often destroys this balance. This is why complexity can be used as a weapon.
It should be noted that the battlefield is fair to any party. In the future battlefield, in order for the enemy to feel the complexity of decision-making in one direction and not to be troubled by the complexity, we must first be superior to the opponent in command and control capabilities. The complexity of battlefield decision-making is mainly reflected in the judgment and decision-making links of the “OODA” loop. Under normal circumstances, the “OODA” cycle can complete the complete cycle from observation, judgment, decision-making to action. However, if there is a way to make the battlefield more complex so that the opponent cannot make effective judgments in a timely manner, and thus cannot enter the decision-making and action links, the opponent’s “OODA” loop can always be limited to the observation and judgment links, and a closed loop cannot be formed. This may be the result of “decision-centered warfare” trying to create complexity. Therefore, how to make quick judgments has become a primary concern. If this cognitive process can be completed quickly with the support of advanced technologies such as artificial intelligence, that is, so-called intelligent cognition can be achieved, the speed of the “OODA” cycle can be greatly accelerated and unilateral advantages can be achieved.
Drawing correct judgments based on observation is the prerequisite for making correct decisions. But this can only be done under the condition of “having cognitive ability”. Currently, in systems such as command information systems and war game deduction systems, these cognitive tasks are basically completed by humans. Past attempts to autonomously complete judgments and decisions by artificial intelligence systems have been almost unsuccessful because the problem of intelligent cognitive modeling has never been solved. The behaviors displayed by various models are more or less “mechanical” and cannot truly show the characteristics of intelligence. Foreign militaries have also been focusing on “human behavior modeling” in recent years, but progress is still slow at present. Why is intelligent cognition so difficult, and what is the difficulty? The author believes that the core difficulty lies in how to understand and deal with complexity.
2. Understand complexity
At the beginning of this century, the Rand Corporation of the United States used a simulation system to conduct more than 1,700 deductions on the combat needs of the U.S. Air Force in response to possible military conflicts in a certain hotspot area around 2005. It then conducted statistical analysis and finally concluded how the U.S. Air Force Conclusion to maintain superiority on the battlefield. This statistical analysis method has a basic assumption: each trial is independent and unordered, and the rules do not affect each other. It’s like tossing a coin. If you toss it heads once, it’s likely to be heads the second time. But if you throw it 10,000 times, the probability of the result being a certain side will get closer to 50%. This method is scientifically accurate when used in physical research, but when transplanted to the study of human social issues such as war, the situation becomes different.
Human beings are cognitive and do not just obey the laws of physics like physical entities. Commanders will not simply repeat mechanically when analyzing combat problems. Normally, when people make decisions, they will consider the previous results, which will lead to adjustments to the next action. In this way, the inherent power law characteristics of human behavior will appear, which is often called the “eight-eighth law”. Therefore, we cannot simply copy physical thinking to think about human society.
The reason for this is mainly because we are often accustomed to thinking about problems in a simple way of reductionism. The simple system structure remains unchanged, the results are deterministic, the cause and effect correspondence is clear, repeatable, predictable, decomposable and reducible, etc., have become our default scientific thinking method. But there are still many complex systems in the world, and these systems have a holistic nature, such as the human body, society, economy, war, etc., all fall into this category. What is the overall nature? That is, what cannot be seen locally, but exists when viewed as a whole, is the overall nature. For example, a living person and a dead person are the same in terms of composition, but the difference between a dead person and a living person lies in whether there is life, and life is a holistic quality. The structure of complex systems is variable and has characteristics such as adaptability, uncertainty, emergence, and nonlinearity, and the results are neither repetitive nor predictable. Society, economy, war, cities, including intelligent systems, these human-related systems all have these characteristics. In fact, they are all typical complex systems. Therefore, war has the characteristics of “no return after victory”, which actually reflects the “unrepeatable” nature of the complex system of war.
It is precisely because of the complexity of complex systems that causes and results cannot correspond one to one, which will lead to the failure of the similarity principle, so it cannot be modeled and studied using traditional methods. In order to solve complex problems, some traditional physics methods were mainly adopted in the past, such as statistical methods and simple life body modeling methods based on Agent. This is the case with the Rand Corporation study mentioned earlier. Although it can solve some problems, when it is used to solve problems related to people, especially cognition, the results obtained deviate greatly from reality and are unsatisfactory. .
Why is this happening? This is because the origins of war complexity and physical complexity are different. The source of physical complexity often lies in the complex laws of physical motion; while the complexity of war comes from human cognition. Because humans are not chaotic particles without thoughts, nor are they lower creatures with simple life logic, but are intelligent creatures with cognitive abilities of judgment and decision-making. People will reflect on the results through causal relationships, sum up experiences and make adjustments, and then decide how to act next. Moreover, human cognition will continue to develop, which will further affect subsequent cognition. However, because cognition is highly uncertain, future actions are difficult to predict.
It can be said that under the current technological conditions, what can be predicted are basically simple systematic laws of the physical world, while the impact of human cognition on society or war is often difficult to predict. Therefore, it is a common mistake we make to use physical thinking to think about human society. Based on the complexity of cognition, which is completely different from those immutable physical laws, when we deal with the complexity of war, we must focus on the core feature of “cognition” and work hard on command and control.
3. Coping with Complexity
The core of “decision-centered warfare” lies in the acceleration of cognition. Because almost all changes in war can be seen as cognitive upgrades and complications. In the author’s opinion, to deal with the “decision-centered battle”, we need to “fight complexity with complexity” and start from the basic work.
The first is to understand the core concept of “decision-centered warfare”. That is to seize the initiative on the battlefield by actively creating complexity. For one’s side, one needs to manage one’s own complexity; for one’s enemy, it means imposing more complexity on the opponent. The second is to understand the changes in the mechanism of war. The evolution speed of combat systems is increasing exponentially, which will make it difficult to perceive, control and manage complex battlefields, and the role of intelligent cognition will become more prominent. To this end, it is necessary to focus on the focus of “command and control” and regard battlefield management capabilities as the key. The third is to find the correct concepts and methods of coping. Starting from war design, we take the direction of decision-making intelligence as a breakthrough.
In recent years, a series of achievements in the field of artificial intelligence have brought hope to solving the problem of intelligent command and decision-making. The AlphaGo series of research has brought breakthroughs to decision-making intelligence technology; and the emergence of the GPT large model has further confirmed that decision-making intelligence and even general artificial intelligence are possible in the future. It now seems that artificial intelligence will be deeply involved in wars in the future, which is a reality that must be faced. And this will bring more complexity to war and battlefields.
Decision intelligence research should be placed at the command and control level. To win a war, command and control decisions need to embody both “science” and “art.” The scientific nature of command and control is mainly reflected in how to do it “when you know how to do it”, such as using the obtained command data (weapons and equipment, force formation, battlefield environment, opponent intelligence, etc.), command methods (tasks, processes, procedures, operations planning, etc.) , planning, optimization, etc.), formulate a combat plan and implement it. The artistry of command and control is reflected in knowing how to do it “when you don’t know how to do it.” This is where true intelligence lies. The method is nothing more than continuous trial and error, accumulating experience, finding ways to solve problems, and forming new scientific knowledge. In fact, in reality, commanders continue to discover and summarize winning rules through trial and error, and each commander also has his own intuition and experience.
Therefore, true intelligence is actually finding solutions to exceptions. Following rules is not intelligence, finding your own way to solve problems is the key. Perhaps this is the core of decision-making intelligence and a goal that requires further efforts.