“There is no place where espionage is not possible” 

–Sun Tsu, around 500 BC [1]

1. Abstract

The Fermi Paradox asks: Given that commonplace assumptions suggest the galaxy should be teeming with intelligent life, why have we not detected any?

Meanwhile, a practically ubiquitous pattern in terrestrial international relations is covert intelligence – i.e. spying on others and keeping information hidden from them for strategic benefit. We seek to understand if such strategic intelligence incentives could explain the Fermi Paradox. 

We break the question into two parts. First, considering game theory and the terrestrial history of spying, we ask we ask to what extent ETI in general would be incentivised to covertly spy on human-level civilisations. Second, we conduct a thought experiment: If ETI were spying on Earth, what would the evidence look like, and is it likely the scientific community would be aware of it?

We find that there are significant and universal incentives to support ETI spying on human-level civilisations for simple, strategic intelligence reasons. We therefore propose this as a solution to the Fermi Paradox. Our analysis suggests ETI spying would likely show predictable patterns of evidence, and that some of this evidence would likely have been detected. But we find that it is unlikely the scientific community in general would be aware, partly due to assumptions of the scientific method causing false negatives with this type of non-reproducible, intentionally deceptive evidence. Finally, we propose how our predictions could be tested. 

2. Introduction

What is the Fermi Paradox?

The Fermi Paradox is simply stated: Current observations suggest that no technologically advanced extraterrestrial intelligence (ETI) has spread through the galaxy; but under commonplace assumptions about galactic civilization formation and expansion, this absence of observation is highly unlikely [2]. Under those assumptions, intelligent life that is thousands or millions of years more advanced than ours should be common, and should have had ample time to spread out to every planetary system in the galaxy, even at humanity’s current speeds of space travel [3]. So where are they?

Recently, the Fermi Paradox has received increased attention after a number of discoveries including that earth-like exoplanets are common [4], which has contributed to a renewed academic interest in the search for ETI [5]. 

Solutions to the Fermi Paradox

Many solutions to the Fermi Paradox have been proposed. It has led some to conclude that humans are the only advanced civilization in the galaxy, either because civilization formation is very rare, or because intelligent civilizations inevitably destroy themselves [3], [6], [7].

Another category of solutions are those in which ETI are indeed common, but due to their behaviour we do not detect them. For example some propose that ETI could not easily get here [2], or would not want to come [3], [8]. Others propose that ETI may be present but typically remain hidden – because we are in a zoo [9], or planetarium of their design [10]. More recently, some have proposed that ETI are common, but for essentially game theoretical reasons typically remain hidden (e.g. due to the risk of being attacked [11], [12]). Such game theoretical approaches to explaining ETI behaviour are appealing because they are potentially universal, explaining why all groups of ETI would have a given behaviour, as opposed to relying on (for example) coincidence that all groups choose not to travel. We build on this type of approach by exploring a related but different game theoretical, behavioural solution to the Fermi Paradox. 

Will ETI be indifferent to detection?

The conclusion that because we have not detected ETI, they do not exist, or would not travel here, contains implicit assumptions about the nature and pattern of ETI. In particular, a common assumption in such arguments is that ETI would be indifferent about use detecting them, and not purposefully try to avoid it [3]. The vast majority of practical research involved in searching for ETI rests on this assumption (e.g. [13], [14]). 

But a closer look at they way terrestrial civilisations interact (and the game theory that dictates it) suggests there may be problems with this assumption. 

In short, spying and secrecy are extremely common.

The conditions that prompt spying and secrecy (i.e. covert intelligence) are defined by game theory, and are very simply satisfied: In any game of imperfect or incomplete information where decisions are made in advance (e.g. rock paper scissors, poker, battleship; business, sport, war, international relations, etc.) there will be an incentive to spy on other players, and keep them from spying on you [15], [16]. For example, consider the children’s game of rock paper scissors: If I have foreknowledge of the strategy you are going to play, I will always win. This immediately creates an incentive for me to spy on you to find out. In turn, this creates a incentive for you to keep information about your strategy secret from me (and vice versa). This dynamic creates an arms race for information (e.g. [17], [18])

similar to those observed in evolutionary biology and elsewhere [19]–[21]. 

The simplicity of the conditions that incentivise covert intelligence contribute to it being widespread. Throughout human history, covert intelligence is practically ubiquitous, has arisen multiple times independently [22], and today is considered a necessary function of government [23]. 

Although the conditions to enable covert intelligence are simple, the subsequent behaviours it incentivises are not. For example, in Rock Paper Scissors, if I do get foreknowledge the move you intend to play but you find out that I know, then my knowledge becomes useless – so there is also an incentive for me to keep my spying on you secret, and for you to feed me false information, etc.. Furthermore, if I know that you know that I know your strategy, then I can still gain advantage – but what I should play depends on whether you know that I know that you know that I know… and so on: I am forced to model infinite hierarchies of knowledge and beliefs [16]. The resultant hall of mirrors is highly prone to deception and is notoriously difficult to navigate [23]. 

The ubiquity and nature of covert intelligence is also often underappreciated. There is a well documented gap in its history and academic literature [24]. For example, the Christian Old Testament (the Jewish Tanakh) contains more references to spies than any history of Britain or of most other countries [22]. This gap has been created by a number of factors, including the inherently secret nature of covert intelligence, a lack of appetite from governments to fund public study of it, and general confusion arising from fictional portrayals [22], [23]. On this latter point David Omand, the former head of the UK’s signals intelligence agency notes “In considering public views of intelligence and security we are dealing with a magical reality, a psychological construct, not direct portrayals of the real world [23]”. 

Secrecy, lack of study and fiction may contribute to why covert intelligence has been neglected as a solution to the Fermi Paradox.

Objective of this paper

Given its ubiquity on Earth, covert intelligence could perhaps be common amongst ETI. If so, this could perhaps explain the Fermi Paradox. 

In this paper we seek to understand if it is plausible that the cause of the Fermi Paradox is due to ETI covert intelligence. To address this ask two questions: 

1. Is it plausible that advanced ETI in general would typically carry out covert intelligence (i.e. spy on, and attempt to remain hidden from) human-level civilisations?

We consider the game theory of covert intelligence, dynamics of knowledge races and common patterns of covert intelligence throughout history, to create a simple qualitative model to understand the conditions under which covert intelligence is likely to happen. We then input basic assumptions about ETI to qualitatively asses their incentives for covert intelligence. 

2. If ETI were carrying out covert intelligence on Earth, would we know?

Next we conduct a thought experiment. We ask if ETI were carrying out covert intelligence against earth, (a) what patterns of evidence would this leave, (b) which of this evidence would be detectable and (c) which of it would identifiable by humanity. To address this question we build on the previous model by considering relevant and common patterns of covert intelligence evidence, detection, identification and methods throughout history. The output constitutes a set of observationally testable predictions for ETI spying.

Results

We find that it is a minimum plausible that ETI will typically remain hidden from, and spy on, civilisations at humanity’s level of development, for basic strategic reasons. This questions the common assumption that ETI will be indifferent about being detected. It poses a solution to the FP that we name the Covert intelligence Hypothesis: 

We also find that in the case of ETI spying, evidence of this activity would have likely been detected – potentially enough to correctly identify the cause as ETI – but that the scientific community in general would be unlikely to be aware, in part due to the deceptive nature of the evidence being largely incompatible with the strict scientific method. We propose a set of observationally testable predictions, and suggest a methodology by which they could be tested. 

Assumptions

This work rests on the following assumptions:

• The solar system not special: intelligent life is common in the galaxy, and much of it is thousands or millions of years older than humanity
• Advanced ETI are rational decision makers (i.e. subject to game theory), commonly develop interstellar travel capabilities, and typically spread out across the galaxy
• We are unaware of ETI

We therefore limit our analysis to ETI expected by these assumptions, i.e. ETI who:

• Are already in the vicinity of the planetary system containing a human-level civilisation (i.e. are either on the planet already, or can travel to it in days / weeks);
• have been in the vicinity for a long time (~thousands of years or more);
• are thousands or millions of years more technologically and scientifically advanced than humanity.

We also allow for the fact that there may be multiple groups of different ETI who have reached any given planetary system.

Is this scientific?

Spying is not a common subject in science. This raises the question: Is this scientific? The key criteria for a scientific hypothesis is falsifiability [25]. All of the analysis in this paper is simply falsifiable: Should any of the above assumptions be disproved then it will fall. It is also verifiable: We present observationally testable predictions, and outline methods for testing. 

However, covert intelligence necessarily involves conspiracy, which is commonly dismissed out of hand in science. Should we dismiss the possibility here, too? The dismissal of conspiracy was originally suggested by Popper [26] – not because conspiracies do not exist, but because they are hard to carry out. Popper’s approach is a heuristic rule, to be applied where it is appropriate. For most of science, where the subject of study is most often realistically incapable of conspiring against the scientist, this is a very useful and time saving approach. (E.g. it is not worth considering the possibility that your odd data on butterfly migrations is because the butterflies are conspiring against you). But in covert intelligence, conspiracies are necessary, and often the default [22]. (E.g. It certainly is worth considering that your odd data on German U-Boat movements is because the German High Command are conspiring against you). Therefore, in this case, Popper’s heuristic is inappropriate: Pre-dismissing conspiracy when it is an extremely unlikely and potentially unfalsifiable possibility (as in most science) is very helpful; Pre-dismissing conspiracy when it is a common and important component of the field is not – and doing so runs a significant risk of type 2 statistical errors (i.e. false negatives).

In a sense, this odd juxtaposition highlights the unique nature of the Fermi Paradox. The existence of ETI is such a vital question when trying to understand the universe that it very clearly aligns with the goals of science, and therefore, whatever the answer to that question is, it must also be worth studying scientifically. But given that we know absolutely nothing about ETI, how can we know if science is an appropriate tool to use to search for it? If we assume ETI will be similar to most subjects that science typically studies (i.e. indifferent or incapable of deceiving the investigator), then science surely is the ideal tool. But what if ETI does not fall into this category, and instead, like most guilty parties in a criminal trial, or child spying on their opponent’s cards, is sophisticated enough, and motivated enough, to present an intentionally deceptive set of data? The scientific method is not configured to cope with such intentional deception (Figure X), which may be why it is not, for example, used to decide guilt in court. Therefore there is a risk we may use an inappropriate tool, and come away confidently (and perhaps incorrectly) saying that ETI does not exist. In the unique case of the Fermi Paradox, an awareness of the limitations of the scientific method may be valuable. 

Finally, a related question is: Would such advanced ETI, even in theory, be detectable? The extent to which covert ETI are detectable obviously depends on their capability. In the case of ETI who are omnipotent with respect to our current capabilities ( we term this “relatively omnipotent”), we would not be able to detect them, and so this case of the hypothesis is untestable. (It is still falsifiable, as described above.) However, to constrain the problem to testable predictions, we limit our analysis to ETI who are not relatively omnipotent. If ETI are indeed common, it may be that there are many groups, and some may fall into either category.

Layout of this paper

In section 3.1 we address the question of the extent to which ETI covert intelligence would be incentivised. We discuss some basic principles of covert intelligence (section 3.1.1), game theory of covert intelligence (section 3.1.2), cost-benefit analysis (section 3.1.3), dynamics of knowledge races (section 3.1.4), basic assumptions about ETI (section 3.1.5), and combine those into an analysis of ETI incentives for covert intelligence (section 3.1.6).

In section 3.2 we address the question of whether the scientific community would generally be aware if ETI were carrying out covert intelligence on Earth. We look at the history of covert intelligence, and discuss common and relevant patterns of evidence (section 3.2.1.), and patterns and methods of detection (section 3.2.2) and identification (section 3.2.3). We then present predictions of what we would expect in the case that ETI were carrying out covert intelligence, focusing on evidence patterns (section 3.2.4), detection patterns (section 3.2.5) and identification patterns (section 3.2.6). We summarise these results in section 3.2.7.

In section 4 we discuss implications, limitations, and how our predictions could be tested, and in section 5 we present concluding remarks. 

3. Models, Results and Predictions

3.1 Would ETI spy?

There is no fully developed framework that describes in detail when and how covert intelligence will happen [27], [28]. Therefore, to assess the potential for ETI covert intelligence we develop a crude qualitative model below, which is informed by general principles of covert intelligence, game theory, the dynamics of arms races / knowledge races, and cost-benefit dynamics.

Some basic principles of intelligence

Firstly, intelligence is not magic, nor is it mythical. It is simply the process of satisfying a basic need for information to reduce ignorance . Rather like any other information gathering and processing activity, intelligence activity involves collecting information about the world and using it to inform your model of the world – to better understand what is happening, why it is happening, and trying to predict what will happen next. The general goals of intelligence are situational awareness and foreknowledge, which provide decision advantage, acting as a force multiplier [22], [23], [27], [29]. 

As briefly discussed in the introduction, perhaps the most confusing parts of covert intelligence are caused by its adversarial nature: You are often trying to predict someone who is at the same time trying to predict you, and so there’s a potential infinite belief and knowledge hierarchy of what you know, what they know, what you think they know, and what you think they know you know, and so on [16]. As a result, there’s an incentive for keeping what you know secret (stealth), and for finding out what they know in secret (spying) [23]. This creates a arms race for information. 

To explore some basic principles of intelligence, we will return to a toy example – playing cards. Although only slightly more complex than Rock Paper Scissors, playing cards has sufficient complexity to demonstrate some key intelligence patterns. (However, in some instances the complexity of the real world generates behaviours that are not obvious in such toy examples, so we will be forced to sometimes consider real world patterns instead). 

Collecting information is often difficult

Imagine you are playing any card game where you have a set of private cards that you can see, but your opponent cannot  (e.g. poker, rummy, bridge, etc.). There is a incentive for you to spy on your opponent to gain foreknowledge of their actions, to help you win. But a key difficulty of collecting secret intelligence is that your opponent is typically seeking to prevent you from accessing their secret information: They are keeping their cards close to their chest. An added difficulty depends on the type of information you are trying to access. In intelligence, information is often divided into capabilities (e.g. what cards do they have?) and intentions (e.g. what do they intend to do with those cards?) [29]. Spying on either type is difficult: 

“There can be no perfect foreknowledge. Secrets are hard to find out, mysteries are hard to assess, and complexities involve second- guessing the moves of one’s own side”. [23]

But spying on intentions is often more difficult, especially as this information often only resides inside your opponent’s head, and can change in an instant: 

“Even the most sophisticated intelligence breaks down when confronted by the whims and vagaries of the changing human mind.” [23]

Despite the ephemeral nature of intentions, they can be extremely strategically important. For example: Are the Russians planning to launch their missiles, or not? 

Secrecy and spying incentivise each other

In your card game, just as you are incentivised to spy on your opponent, you are also incentivised to keep your cards secret from them, and repel their efforts to spy on you. But even this effort of secrecy incentivises spying, because to stay secret, it very much helps to know your opponent’s capabilities of spying, so that you can mitigate them – which often requires spying on them to find out [17], [29]. Similarly, there will be an incentive to keep your efforts to spy on your opponent secret – even ignoring any punishment, it is simply easier to spy on someone if they are unaware that you are doing it; also, when spying on intentions, information is often only valuable if your opponent is unaware that you know it. 

Secrecy is only worthwhile for information that is not known to the other side.

Another obvious point is that it is only worth trying to keep the cards hidden that are not publicly known (in game theory terminology: common knowledge [16]). For example in some forms of Poker there are public cards (face up) that are known to everyone, and private cards for each player. This explains why for example in the modern world spying and secrecy only occurs on a small subset of information – not because spying is an inherently marginal activity, but because most information is common knowledge (e.g. there is no benefit for the US to try to keep wheels, steam engines or or the location of South America from China, because they are publicly known). 

In short, the amount of private information defines how much scope there is for secrecy and spying. If there is no private information, then there is no benefit to secrecy at all. And vice versa, if all of your information is private (as in many card games), then complete secrecy is incentivised. 

Some private information is more valuable than others

When playing cards, it is often the case that learning of some of your opponents cards would be more valuable than others. For example in Poker, knowing that your opponent has the one card that could beat your hand is much more valuable than knowing which other cards they hold. In general, the more potential strategic loss information would cause if it fell into the hands of your opponent, or strategic gain it would give if captured, the more valuable that information, and hence the greater the incentive for secrecy or spying. 

This is hard to generalise, but pieces of information that would cause very large scale impact if known (real world examples might include a recent discovery for workable nuclear fusion, how to break widely used cryptography, etc.) are likely to be of higher strategic value than those with low level impact (e.g. the engineering specifications of a now defunct technology). 

Arms race dynamics

The incentives of covert intelligence create a arms race for information in which what matters is the marginal amount of advantage over your competitor, and not your absolute capability. You may have highly sophisticated methods of spying on your opponent in cards, but if their methods of secrecy are more advanced, your efforts will probably count for nothing. As such, a typical pattern of covert intelligence (arms races generally) is that of tit-for-tat improvements in response to improvements of the other side (e.g. improvements of detector capabilities of one side in response to improved stealth of the other) [17], [19], [21]. 

There is often an incentive to spy even without a strategic target

Let us imagine your have been playing cards against your opponent, and each spying on the other, for some time. You know how your opponent spies on you, and you have methods to effectively counteract it. You might therefore be tempted to stop spying on how they spy on you, because you already know the answer, and it takes constant effort. But in this case, how will you know if they start doing something different? Worse, what if they start doing something entirely new that you didn’t expect, like conspiring with another opponent? 

The nature of having incomplete and imperfect information means that you don’t know what you don’t know (and so can be helpful to try to find out things, even if you don’t know what you’re looking for). This was famously articulated by Donald Rumsfeld: “There are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns—the ones we don’t know we don’t know. And if one looks throughout the history of our country and other free countries, it is the latter category that tends to be the difficult ones”. [30] The risk of change, and unknown unknowns, provide a common incentive to spy even in the absence of known strategic target [16], [29].

Mostly, everyone spies on everyone, always

In a similar way, in the Art of War Sun Tsu explains that intelligence is not just for war, but is highly valuable in peacetime, as an investment in early warning in case of a threat or opportunity [1], [29]. Spying is also not reserved for powerful enemies (in the modern world, practically all countries spy on all others, including their allies [22]); neither is it purely about short term solutions:

[Governments use intelligence] not just to be able to make predictive judgments about the future course of events, but to realize what the world would then look like so as to be able to identify and start to implement policies and partnerships now that would mitigate the risks to society of such trends. 

Although Chess contains no private information, it is perhaps a more relevant toy example in this case than playing cards, as like international relations, it is highly strategic and contains many interdependencies. To become a chess grandmaster, it is vital to consider the potential moves and strategic impact of all of the pieces (including those that that currently pose no direct threat or opportunity) –not just the most powerful, or currently important. Many chess games have hinged on the actions of a single pawn, just as many seemingly unimportant countries have at times become highly strategically valuable [22], and so a common approach is to collect intelligence on all of them.

Technology races and covert action

Not all covert activity or intelligence is about decision advantage. For example, 

commonly the target of intelligence gathering is technology (e.g. how to make nuclear weapons, or advanced stealth, etc.) [22]. Such information can simply provide capabilities advantage, as opposed to decision advantage. In addition, some covert activity is less about gathering intelligence but more about covert action – i.e. actively attempting to change the course of events by unattributable action [29]. This falls under the same general umbrella of covert actions incentivised for simple strategic reasons. 

Covert intelligence appears to be an emergent behaviour

Given that a simple card game can incentivise covert intelligence, it is perhaps not surprising that is appears to be an emergent behaviour in the history of civilisation. Occasionally in the past, some states have, for various reasons, shunned covert intelligence. This has mostly happened either where states are so large and powerful they can win battles with brute force alone, without the force multiplier that intelligence provides (e.g. China during the 18th-19th centuries), or where for ideological reasons, covert intelligence is seen as distasteful (e.g. in ancient Athens). In all of these recorded cases, following some disastrous losses, covert intelligence was reinstated [22]. 

The Game Theory of covert intelligence

Game theory is the study of mathematical models of strategic interaction among rational decision-makers. It has been widely used to successfully predict the dynamics of social system across different types and substrates. [16]

There is relatively little work on the game theory of intelligence, but models have shown that the requirements for covert intelligence are very simple: It is incentivised in games of imperfect or incomplete information, and is subject to cost / benefit calculations (e.g. [15]). Imperfect information games are those where players have only partial knowledge about the current state of the game, such as Poker or Battleship, and incomplete information games are those where players have only partial knowledge about the opponents utility function, such as auctions. Most real world situations fall into both of these categories – for example, sport, business, politics, war, international relations, etc. [16], [29].

Although games are inherently cooperative, they do not always result in hostility. Many games spontaneously produce highly cooperative behaviour to reach mutually beneficial solutions, and game theory has been a central tool in understanding how cooperation can happen [31], [32]. In a similar way, the results of covert intelligence is not always hostile, and can result in mutually beneficial solutions for both sides (e.g. giving advance warning of disagreements and giving time for diplomatic solutions.) [23]

Cost / Benefit Analysis of covert intelligence

There is an economic aspect of covert intelligence. Even if covert intelligence is incentivised, if the costs outweigh the benefits it will not be the rational decision. We outline some of the simple factors affecting the cost / benefit below.

Benefits

We have already discussed the benefits of covert intelligence at length. In general, the bigger the potential strategic benefit of covert intelligence, the greater cost can be justified. 

Because with ETI we are considering a highly advanced opponent, a relevant factor that affects the level of benefit is knowledge asymmetry. The greater knowledge asymmetry in your favour, the more decision or capabilities advantage you have to lose, and hence in general the more incentive there is for you to maintain that asymmetry – i.e. the greater the incentive for secrecy. (This likely also increases your incentive to spy on your opponent in a defensive manner to understand and defend against their attempts to spy on you.) In the same way, the greater knowledge asymmetry in your opponent’s favour, the more potential benefit there is from removing that asymmetry – i.e. the greater the incentive for spying  (and also to keep that spying, and your capabilities for spying, secret) [33]. 

Costs

Two key contributors to the cost of covert intelligence are risks and capabilities. Risks of secret intelligence include outcomes such as punishment from being caught, exposing your capabilities (etc.). The greater the risk, the greater the potential cost. 

Capabilities refer to your own capability at conducting covert intelligence operations. They include logistics, stealth, detection, modelling, analysis, and the level of knowledge you have about your opponent. In general, the greater capabilities asymmetry in your favour, the lower the cost of covert intelligence, and vice versa. For example, the US’s cost of spying on Afghani opium farmers without being detected is probably relatively small, because the US already have highly advanced satellite and drone technology than can simply be redeployed to the area. Meanwhile, the farmers may only have only rudimentary tools to detect the spying, and limited logistical capabilities to react to it – making secrecy a relatively low cost task for the US, given their existing capabilities. On the other hand, although there would be potentially a huge amount to gain for Afghani farmers to access the military and commercial secrets of the US, the huge capabilities asymmetry against them in terms of accessing, understanding and using that information may mean that the cost of success is probably prohibitive. 

Capabilities constrain the scale of SS

Capabilities also partly constrain the scale of covert intelligence possible. For example, in the 19th century is was generally thought that only tactical (i.e. small scale) surprise was possible in war, because of the lack of logistic capability to mount a strategic (i.e. large scale) offensive. However, with the industrialisation of war and invention of the railway, telegraph (etc.), the scale of possible deception increased dramatically, and strategic surprise became possible [23]. In general, greater capabilities enable larger scale operations. 

How do we use this to tell if ETI will spy?

In table 1, we collect the qualitative factors described above to assess the plausibility of ETI spying. We assess a known positive case (modern day US / China international relations) and two known negative cases (human / chimpanzee interactions, and human / C. Elegans interactions.)

Basic assumptions about ETI

To use as an input to the model, we make the following basic assumptions about ETI. These reflect the general scope of the paper stated in the introduction. (For simplicity we describe the interaction as ETI / Humanity, but the assumptions are mostly general, and so the results may apply generally to interactions between ETI  and Human level civilizations.) 

ETI are:

• Capable of building technology to enable interstellar travel to earth
• Thousands or millions of years more technologically advanced than us
• Are already in the vicinity of earth, and have been for thousands of years

And

• We are unaware of them

Results