Symbionts or Parasites: Our Future with AI
Human Agency at the Crossroads of AI Partnership and Dependence.
Now that the debate over artificial intelligence — and its dangers and benefits for humanity — is in full force, I return to the question posed in my book Symbionts or Parasites? Humanity’s Crossroads: Will intelligent technologies become partners that expand human freedom, or systems on which we depend at the cost of our autonomy? [1]
The answer cannot be reduced to predictions about job losses or productivity. Technology reshapes habits, language, institutions, and expectations. As Marshall McLuhan argued, media extend human capacities while transforming the people and societies that use them [2]. AI is the newest and perhaps most consequential extension.
The question, then, is not simply what AI can do. It is what kinds of relationships, incentives, and human beings its use will produce. J. C. R. Licklider imagined human-computer symbiosis as a partnership in which each side contributes what it does best [3]. My earlier work on symbiogenetic human digitization argues that prolonged interaction with digital systems can make technology part of how people perceive, communicate, and adapt, creating both new capacities and new forms of subordination [4]. That possibility remains open — but it is not guaranteed.
The Symbiotic Possibility
A healthy symbiosis combines different strengths. AI can process immense quantities of information, detect patterns, translate languages, and automate repetitive work. Human beings contribute judgment, empathy, responsibility, imagination, and contextual understanding. Research on collaborative intelligence argues that organizations often gain more when people and machines augment one another than when automation merely displaces workers [5].
This partnership can improve medicine, scientific research, accessibility, public services, and education. Yet benefits depend on design. AI should help clinicians interpret evidence without reducing patients to data profiles; help teachers adapt instruction without replacing the relationship at the heart of learning; and help workers perform better without stripping them of discretion, dignity, or opportunities to develop expertise.
A further opportunity is cognitive diversity. Used carefully, AI can challenge assumptions, simulate competing explanations, and expose users to perspectives they might overlook. Used lazily, however, it can produce intellectual uniformity: millions of people drawing from similar systems, phrasings, rankings, and hidden defaults. Symbiosis therefore requires active human disagreement, not passive acceptance.
The Parasitic Danger
The relationship becomes parasitic when convenience systematically erodes autonomy. In biology, parasitism is defined not merely by resource extraction, but by host manipulation — the parasite altering the host’s behavioral patterns to sustain its own reproductive cycle. In digital ecosystems, platforms execute a parallel maneuver: capturing human attention, rewarding compulsive engagement, and converting behavioral surplus into predictive instruments of influence. Shoshana Zuboff conceptualizes this economic imperative as surveillance capitalism, while Daniel Susser, Beate Roessler, and Helen Nissenbaum warn that data-driven architectures covertly manipulate human choices, obscuring the intervention from the subject [6, 7]. My analysis of algorithmic deception demonstrates that platforms do not invent falsehood; rather, they industrialize it by prioritizing virality, emotional stimulation, and tribal affirmation [8]. Generative AI accelerates this dynamic, rendering influence conversational, deeply personalized, and exceptionally difficult to inspect or contest.
The deeper hazard is cognitive surrender. When we delegate memory, composition, contextual interpretation, and ethical deliberation to automated systems, mechanical efficiency expands while autonomous human capacity atrophies. Research on transactive memory illustrates that individuals retain less substantive information when they expect digital repositories to store it, preserving only the retrieval pathway [9]. My work on language and netsocieties reveals that these digital environments do not merely transmit information; they actively reconstruct identity, discourse, and social practices, transforming cognitive shortcuts into entrenched cultural dependencies [10]. Assistance metastasizes into parasitism when the tool ceases to extend human judgment and instead substitutes for it — eroding the productive cognitive friction necessary to cultivate independent intellect.
Moral Agency vs. Algorithmic Mediation
Humans are not inherently evil, but AI can amplify both our virtues and our destructive impulses. The central question is not whether AI itself is evil, but whether our institutions, incentives, and moral judgment can prevent it from becoming a tool for manipulation, exploitation, surveillance, deception, or violence.
Every powerful technology inherits the contradictions of its creators. AI can broaden access to knowledge, diagnose diseases, assist people with disabilities, and help societies respond to complex problems. The same capabilities can be redirected toward profiling, automated discrimination, persuasive propaganda, fraud, or coercion. AI does not supply the moral purpose; people and institutions do.
For that reason, responsibility cannot be displaced onto the machine. Developers choose objectives, companies establish incentives, governments determine permissible uses, and citizens decide what forms of convenience they will accept. The danger begins when each participant claims that the system — not the human chain behind it — made the decision.
Human Judgment Must Remain Central
AI is not inherently good or evil. Its trajectory reflects human intent, market incentives, regulatory frameworks, training corpora, and institutional priorities. The essential questions are therefore fundamentally political and social: Who designs the system? Whose values govern its objectives? Who has the authority to audit it? Who captures its economic gains, and who bears its social costs? As Cathy O’Neil has demonstrated, opaque algorithmic systems regularly entrench historic inequities under an aesthetic veneer of mathematical neutrality [11].
Human judgment cannot be dismissed as an inefficient friction in the path of total automation; it is the essential anchor of moral and legal accountability. An algorithmic model cannot bear moral culpability, nor can it understand why efficiency must sometimes yield to equity, procedural fairness, privacy, or human liberty. Meaningful human oversight remains central to international frameworks, such as UNESCO’s Recommendation on the Ethics of Artificial Intelligence [12]. Delegating operational execution to an algorithm does not dissolve the human duty behind the outcome.
Consequently, digital and AI literacy must be reimagined as critical civic defense rather than mere vocational upskilling. Citizens require the discernment to cross-examine automated outputs, verify algorithmic hallucinations against evidentiary baselines, interrogate synthetic media, and know when to refuse mechanical delegation altogether. My empirical examination of graduates during the COVID-19 pandemic found that digital literacy correlates directly with socioeconomic agency and highlighted persistent structural disparities in access, technical equity, and institutional support [13]. In educational institutions, this necessitates defending what learning theorists’ term “desirable difficulties the deliberate friction through which students forge critical synthesis, memory, and personal voice. Subcontracting every nascent idea to a generative model may polish the final artifact, but it impoverishes the thinker.
Ultimately, governance must confront the asymmetric concentration of capital and infrastructure. A small oligopoly of technology corporations exercises gatekeeping power over the foundational cloud-computing resources, proprietary datasets, and specialized talent required to deploy frontier models. This bottleneck allows private commercial entities to unilaterally police labor markets, structure public knowledge, and delineate permissible public speech. My research on algorithmic governance argues that accountability cannot fixate solely on post hoc moderation of isolated outputs; it must enforce systemic obligations, including statutory duties of care and safety-by-design architectures that treat automated curation as active amplification rather than passive conduit carriage [14]. Furthermore, proprietary secrecy and algorithmic opacity distribute institutional culpability so broadly that automated decisions become impossible to appeal [15].
Get Dr. Pedro A. González Jr MJ MAC’s stories in your inbox
Join Medium for free to get updates from this writer.
Robust governance must therefore extend beyond voluntary corporate self-regulation. It requires structural interventions: antitrust scrutiny to prevent vertical infrastructural foreclosure, publicly funded compute commons to preserve open academic inquiry, independent third-party audits with binding inspection powers, and enforceable rights to human appeal. Crucially, governance must also reckon with the material extraction underlying artificial intelligence: the immense energy demands of hyperscale data centers, which strain municipal power grids, deplete regional water resources for cooling, and exacerbate environmental degradation. As the NIST AI Risk Management Framework emphasizes, responsible governance must encompass the full life cycle of the system — from mineral and ecological costs to training, deployment, and downstream socio-technical impact [16].
At Humanity’s Crossroads
We stand at a genuine crossroads. AI may help create a more capable, inclusive, and knowledgeable society. It may also magnify inequality, weaken privacy, standardize thought, intensify surveillance, and make dependency appear inevitable. Neither outcome is dictated by the technology itself.
We still have the power to choose the terms of the relationship. That means welcoming innovation without surrendering judgment, using automation without abandoning responsibility, and accepting assistance without confusing it with wisdom. It also means judging progress by more than speed and profit: the proper measures are whether technology enlarges human agency, distributes benefits fairly, and strengthens our capacity to live together.
Computational optimization can identify efficient means toward specified ends, but prudence — or phronesis — is the distinctly human capacity to judge which ends are worthy, just, and appropriate in circumstances.
Agency must therefore be asserted simultaneously across the pedagogical, legislative, and civic spheres: individuals must cultivate independent judgment, demand accountable laws, and participate actively in shaping institutions and the state, through which collective choices about AI are made.
The real test of artificial intelligence will not be whether machines can imitate human beings. It will be a matter of whether human beings can preserve autonomy, dignity, pluralism, and moral purpose while using machines of unprecedented power. If we succeed, AI can become a genuine symbiont. If we fail, the parasite may not be the machine alone, but the habits, markets, and institutions of dependence we allowed to grow around it.
References
[1] González, Jr., P. A. (2023a). Symbionts or Parasites? Humanity’s Crossroads. Editorial Letra Viva.
[2] McLuhan, M. (1964). Understanding Media: The Extensions of Man. McGraw-Hill.
[3] Licklider, J. C. R. (1960). “Man-Computer Symbiosis.” IRE Transactions on Human Factors in Electronics, HFE-1, 4–11.
[4] González, Jr., P. A. (2023b). The Symbiogenetic Human Digitization: How the Interaction with the Internet Through Digital Technologies Created a Symbiogenetic Human Digitization. Editorial Letra Viva.
[5] Wilson, H. J., & Daugherty, P. R. (2018). Collaborative intelligence: Humans and AI are joining forces. Harvard Business Review, 96(4), 114–123.
[6] Zuboff, S. (2019). The Age of Surveillance Capitalism. PublicAffairs.
[7] Susser, D., Roessler, B., & Nissenbaum, H. (2019). Technology, autonomy, and manipulation. Internet Policy Review, 8(2).
[8] González, Jr., P. A. (2026a). The architecture of deception. Medium.
[9] Sparrow, B., Liu, J., & Wegner, D. M. (2011). Google effects on memory. Science, 333(6043), 776–778.
[10] González, Jr., P. A. (2021). Our Demons on the Machine: Language, Netsocieties and Illiteracy on the Web. Editorial Letra Viva.
[11] O’Neil, C. (2016). Weapons of Math Destruction. Crown.
[12] UNESCO. (2021). Recommendation on the Ethics of Artificial Intelligence.
[13] González, Jr., P. A. (2023c). Digital Literacy in Employability of Graduates During COVID-19: A Qualitative Phenomenological Study (Doctoral dissertation, St. Thomas University). ProQuest LLC.
[14] González, Jr., P. A. (2026b). Digital Dictatorship and the Law: Regulating Algorithmic Governance After Section 230. Academia.
[15] González, Jr., P. A. (2026c). The architecture of algorithmic secrecy: Governing through opacity — AI, secrecy, and state power. Academia.
[16] National Institute of Standards and Technology. (2023). Artificial Intelligence Risk
