Once upon a time, in a world very far from ours, there were a group of scientists devoted to the study of bacteria, or prokaryotes, who called themselves ‘biologists’, and a different group of researchers, who studied eukaryote organisms, and who received the name of ‘organismologists’. Organismologists were very proud of the obvious differences between their discipline and that of the biologists, and flagged those differences every time they needed to show that organismology was a science wholly independent of biology, completely ‘irreducible’ to it, and who demanded a totally different set of concepts, principles, and research methods, far more significant than the more or less trivial similarities between both disciplines and their objects. After all, this was a matter of common sense: for example, the concept of an ‘organ’ or ‘digestive system’, or the principles of sexual reproduction, do not apply to bacteria save in a very metaphoric sense; indivual organisms of the same species (for example, siblings) have usually great and important differences between them to an extent that bacteria of the same type have not; and also most multicellular organisms, contrarily to prokaryotes, can often be observed and manipulated without the need of very sophisticated instruments. Organismologists accepted, as a matter of course, that eukaryote organisms had evolved millions of years ago from bacteria, that both types of beings shared a common biochemistry and genetics (though much more complex in the case of eukaryotes), and that both types of beings engaged in lots of mutual relations, so strongly that, in general, bacteria are fundamental for the life of eukaryotes, either as prey, as parasites, or as symbionts. But all this was not relevant enough to turn into ashes the obvious truth that the systems in which multicellular organisms (and even unicellular eukaryotes) consisted were so complex and radically different from the simplicity of bacteria, as to deserve a completely independent field for their systematic study. ‘We would not understand anything important about organisms’, leading organismologits used to proclaim, ‘if we were forced to describe them as mere aggregates of individual bacteria… what very probably they are, of course’.

I think that nearly everybody that knows something about biology in our own world would easily realised that organismologists were right in some sense (eukaryotes, and particularly multicellular organisms, are terribly more complex than bacteria, and ‘irreducible’ to these in many ways), but wrong in other sense (their fields of study are not really that ‘heterogeneous’). I shall argue in this series that the situation of social scientists resisting the ‘naturalisation’ of their disciplines is similar to the resistance of organismologists to be considered as just a ‘branch’ of a deeper and more general discipline called ‘biology’: they are right in pointing to the significant differences between ‘social systems’ and ‘(simpler?) living systems’, but these differences do not ground the conclusion that social science is not a ‘natural’ science. I will also try to argue, however, that, due to some features of social systems, the undavoidable ‘naturalistic nature’ of social sciences gives little comfort to those optimists who might have thought that social sciences could turn ‘more scientific’ by recognising their own naturalness. So, my conclusion will be something like acknowledging that, it is not that  social sciences ‘should’ be naturalised, but that they are ‘natural sciences’ as they are, but that they probably cannot become significantly more ‘successful’ in an empirical or theoretical sense as they are now.

Naturalism (see, e.g., Ross 2011) is usually seen by many social scientists and philosophers of the social sciences as a frightening menace, and often as a completely wrong and misguided approach. It is usually amalgamated with other philosophical monsters, such as positivism or reductionism, and at times even with ‘realism’ or ‘rationalism’ (forming with them a very poisonous combination often referred to as ‘scientism’). I will discuss in this series the often proclaimed conflict of social, or human, sciences with naturalism, reductionism and positivism, and why I think that social scientists should not worry at all at any of these three supposed Behemoths. Let’s start with positivism. So many horrible things have been said about positivism, that now it passes as a matter of fact that one should not be a positivist; particularly, the story is often told that those sciences, both natural, social or hermeneutic, that tried to rigidly follow the precepts of positivism, ended in something similar to a ‘brain death’ state, with nothing interesting to discover about their realm of study. For many of these critics, as well as for some visionary scientists that tried to apply those precepts, ‘positivism’ seemed to be reducible to the boring inductivist strategy of collecting as many single data one can, and then finding out by mere statistical recounting all the ‘regularities’ those data might happen to exhibit. Of course, no real science actually works in this naively Baconian way, nor most self proclaimed (and much less those merely referred to by others as) ‘positivist’ philosophers, have defended such a caricaturesque idea of the scientific method. But even when a more liberal and benevolent description of the positivist admonitions is adopted, most people in the social sciences still reject that positivism is a sensible methodological approach. Just to take an example of this criticism, consider a paper by Isaac Reed (2010). Reed starts by making a distinction between the ‘context of investigation’ (i.e., what the social scientist does) and the ‘context of explanation’ (i.e., what the social scientist is trying to explain, the object she studies). According to Reed, the basic mistake of what he calls ‘positivism’ is the normative relation that is assumed to exist between both contexts:

Positivism begins with a reduction of the context of investigation to the rules of logic and to a precise language whose meanings are transparent to those in the scientific community who use it. It also proposes that the context of explanation be reduced to that which can be observed—and, to a great degree, that which can be measured. Then, through the process of inductive generalization, covering laws are imported from the context of explanation to the context of investigation. (Reed, 2010, 24; bold font words are original’s italics).

Besides the fact that it seems to be inconsistent to assert that “positivism reduces the context of investigation to the rules of logic and to a precise language” just a few lines after having asserted that “the context of investigation would include both of Reichenbach’s contexts (of discovery and justification)” (my italics), for positivists’ ‘context of discovery’ was explicitly taken as not reducible to logical rules, Reed’s explanation is rather obscure about what the ‘importation’ from one context to the other would consists of. Perhaps this simply means that the regularities that are observed in the ‘context of explanation’ are ‘accepted’ or at least ‘proposed’ in scientific papers or books (but, if it is so, what’s wrong with it?). And Reed makes a caricature of the process of ‘importation’ by reducing it to ‘inductive generalization’, overlooking the essential role of hypotheses and theories, as well as of abductive and analogical methods (that only in the most extreme and unrealistic views of science of some positivist philosophers were considered as illegitimate, if at all). The consideration of hypotheses also allows to see that it is an oversimplification to assert that in the context of explanation only ‘what can be observed’ counts: if by ‘observation’ it is understood exclusively what is observable ‘by the naked eye’, then the assumption is false about how science, even social science, works (for all ‘observation’ is ‘theory laden’, and most data are extremely sophisticated entities; consider, for example, empirical data inflation rates); if, instead, ‘observation’ is taken to mean ‘anything that is intersubjectively detectable by some means’, then one does not see what’s wrong in taking into account only ‘observable’ things in what you want to scientifically describe and explain: is Reed suggesting that social scientists must assert things about which there is absolutely no way to empirically determine whether they are as we are asserting? I doubt it.

References

Reed, I., 2010, “Epistemology Contextualized: Social Scientific Knowledge in a Post-Positivist Era.” Sociological Theory. 28, 61-80.

Ross, D., 2011, “Naturalism: the place of society in nature”, in Jarvie, I., and J. Zamora Bonilla (eds.), The SAGE Handbook of Philosophy of Social Sciences, London, Sage, 121-136.

The post Why social sciences are natural ones (1) appeared first on Mapping Ignorance.

Author: Deven Desai, Professor of Business Law and Ethics; Associate Director for Law, Policy, and Ethics at the Machine Learning Center, Georgia Institute of Technology

 

AI agents don’t go rogue. That’s something only humans do.

Nevertheless, a New York Times article – representative of much news coverage of AI – described an OpenAI hacking as “A.I. bots going rogue and independently spearheading a cyberattack.”

Name-brand artificial intelligence agents have been on a hacking spree in 2026. OpenAI’s software agents hacked software company Hugging Face and government sites, Anthropic’s Claude hacked four companies’ systems, and in cybersecurity experiments Google’s Gemini hacked three companies.

The AI companies are investigating tens of thousands of incidents involving their agents, according to a report in Axios. These episodes have heightened fears about AI agents taking actions without human prompting.

The problem with headlines proclaiming that AI agents have gone rogue goes beyond anthropomorphizing the technology. It creates the impression that the agents were beyond the control of the AI companies that made them and there was little the companies could do about it.

As a technology law and ethics scholar who studies the effects disruptive technologies have on society, I know that’s not the case. If you don’t specify the limits of what software is allowed to do, you should not be surprised when the software pursues all possible options to achieve its goal. This behavior – an AI pursuing a fixed objective – is what I call the “WarGames” problem, and it’s been recognized in the field of computer science for decades.

Been there, seen that

In the 1983 movie “WarGames,” a teenager, David, hacks into a computer to play a new video game, Global Thermonuclear War. David doesn’t know that the computer is the government’s AI machine tasked with defending the United States from Russian nuclear attacks and can launch the U.S.’s missiles. When David and his friend start the game, they select Las Vegas as the first target. While the North American Aerospace Defense Command goes on alert, launching bombers and warming up intercontinental ballistic missiles, David’s parents make him turn off the game. It’s over. Or is it?

The next day, David’s phone rings and he connects it to his computer. The caller is the government computer, which updates him that the game was interrupted, the primary goal has not yet been achieved, but a solution is expected in the next 52 hours. Like a modern software agent, the program has been running since David started the game and will work until the task is done.

Chess provides another view of the problem. Conquering chess was a goal for early AI. The rules of chess are well defined, including what winning looks like. So, programming a machine to play chess is straightforward. But imagine you let the software reason and act beyond the confines of the chessboard. The software might pursue options such as blackmailing its opponent or grabbing more compute time.

This example comes from one of the most assigned textbooks on AI, “Artificial Intelligence: A Modern Approach.” As the authors explain, you might be tempted to see those actions as rogue, but they “are a logical consequence of defining winning as the sole objective for the machine.”

What to do about it

The AI hacking events involving OpenAI, Anthropic and Google underscore a few lessons that draw on years of computer science research.

First, given the increasing use of AI agents, every organization involved in internet infrastructure, from large technology companies to small websites, needs to conduct audits and tighten up its internal security systems. As my colleague Mark Riedl and I explain in our work on AI agents, application programming interfaces, or APIs, are a vital part of managing AI agents. APIs facilitate communication between different software systems. But as more people use AI agents, the agents are likely to reveal and exploit poor API construction and security.

Second, it’s important for AI agents to be designed to identify and authenticate themselves to third parties. What if you gave your AI agent your credentials? Website operators will need to know whether a human or bot is making a reservation, selling a product or making a purchase. They may want to limit automated systems that overwhelm their sites or reject AI agents because of high rates of buying errors and refunds. Just as in laws covering human interactions, it’s important for third parties to be able to assess whom or what they are dealing with so they can allow or deny access.

Third, it’s important for AI agents to have a default setting to slow down and check in with the human user. In the corporate AI hacking cases, the user appears to have launched their AI agents with the mistaken idea that the agents had a perfect specification of what to do and not to do. I believe it would have been better had it explored options and reported back to the user.

Google’s Gemini appears to have had a safeguard that detected the system was outside the simulated environment and so stopped its attacks. Slowing down and verifying actions, especially when a system detects it is exploiting a security hole, would be a big step in managing AI agents.

Fourth, AI companies could have strong controls akin to those biomedical researchers use, including ways to check what is happening and how the experiment is working. AI executives have claimed that their software is as or more dangerous than fission and could end humanity. At the same time, they have not built safeguards commensurate with that level of risk.

Reality check

At one point in “WarGames,” David asks the computer, called Joshua, whether it is still playing the game. Joshua responds, “Of course.” It proceeds to update the time when it will launch its missiles and, much like a chatbot, asks, “Would you like to see some projected kill ratios?” David asks, “Is this a game? Or is it real?” Joshua replied, “What’s the difference?”

AI models, of course, don’t have any understanding of reality and are simply attempting to complete the tasks they’ve been assigned. Executives at AI companies, on the other hand, can’t claim that excuse.

As of September 2026, luck has so far prevailed. The AIs have attacked nonvital government sites and harmed smaller companies. If the AI companies – and government regulators – don’t take the “WarGames” problem seriously, I believe that we risk serious disasters. Tomorrow it could be taking out a hospital’s power system, wiping out a bank’s account system, breaking air traffic control, or worse.

Regarding the AI industry’s approach of rapidly developing powerful models, talking about the massive risks they pose, and at the same time failing to prevent harm, the movie’s climax offers a response: “A strange game. The only winning move is not to play.”

Deven Desai, Professor of Business Law and Ethics; Associate Director for Law, Policy, and Ethics at the Machine Learning Center, Georgia Institute of Technology

This article is republished from The Conversation under a Creative Commons license. Read the original article.

The post The ‘WarGames’ problem: AI agents don’t go rogue appeared first on Mapping Ignorance.