<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Marinamistura on Medium]]></title>
        <description><![CDATA[Stories by Marinamistura on Medium]]></description>
        <link>https://medium.com/@marinamistura?source=rss-6eb45f2f1253------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*uL5K1ArotDt6_FeL</url>
            <title>Stories by Marinamistura on Medium</title>
            <link>https://medium.com/@marinamistura?source=rss-6eb45f2f1253------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 15:33:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@marinamistura/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[When Reality Is No Longer Shared: Hyperpersonalization as AI's Next Challenge]]></title>
            <link>https://medium.com/@marinamistura/when-reality-is-no-longer-shared-hyperpersonalization-as-ais-next-challenge-b546d491a5f0?source=rss-6eb45f2f1253------2</link>
            <guid isPermaLink="false">https://medium.com/p/b546d491a5f0</guid>
            <category><![CDATA[science]]></category>
            <category><![CDATA[psychology]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Marinamistura]]></dc:creator>
            <pubDate>Wed, 08 Jul 2026 23:32:40 GMT</pubDate>
            <atom:updated>2026-07-08T23:32:40.056Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>When Reality Is No Longer Shared: Hyperpersonalization as AI&#39;s Next Challenge</strong></p><p>News organizations have always had editorial lines. Information has never been consumed free of bias. What is changing today, however, is not simply the presence of bias, but its nature and scale.</p><p>When we open a digital newspaper, the front page is not necessarily the same for every reader. This is not inherently deceptive: the order, visual hierarchy, and selection of stories are often personalized to make the content more relevant to each individual. Yet the way information is organized inevitably shapes how we perceive reality.</p><p>We are living in a time in which conspiracy narratives increasingly coexist with superficial or fragmented interpretations of scientific ideas, giving rise to beliefs such as the possibility that we are living inside a simulation. The purpose of this article is not to argue for or against such claims. Instead, it is to reflect on the technological conditions that may contribute to their spread.</p><p>Until recently, algorithms selected which content to show us from an existing pool of information. With generative AI, content no longer needs to be selected—it can be created specifically for each individual. Its tone, examples, sequence of arguments, length, analogies, and even the way it anticipates and responds to objections can be dynamically tailored to a single user.</p><p>This represents a profound shift.</p><p>The better content aligns with a person&#39;s preferences, beliefs, and expectations, the greater their satisfaction, engagement, and time spent on the platform. From a business perspective, hyperpersonalization is an extraordinarily successful strategy, making it unlikely that companies developing these systems will abandon it voluntarily.</p><p>For much of the twentieth century, mass media addressed millions of people with broadly shared narratives. Political disagreements certainly existed, but there remained a common public reality—a shared set of events and references that different groups could interpret differently while still acknowledging the same underlying facts.</p><p>Social media began to fragment that shared space. Communities formed around increasingly specific interests, creating strong social bonds even within relatively small groups.</p><p>Generative AI may mark the beginning of a new stage. We are no longer speaking merely about communities, but about potentially individual narratives. Each person can interact with an information environment designed exclusively for them, optimized to capture their attention and reinforce what they are already inclined to believe.</p><p>The challenge, therefore, extends far beyond fake news, synthetic images, or influencers presenting carefully constructed versions of reality. The deeper issue is the possibility that every individual comes to inhabit an informational world that becomes progressively more difficult to compare with anyone else&#39;s.</p><p>Once the experience of a shared reality begins to disappear, confirmation bias is no longer a side effect of our information environment—it becomes its organizing principle.</p><p>This is no longer Descartes&#39; methodical doubt, conceived as a philosophical exercise to establish more reliable knowledge. Instead, doubt is increasingly accompanied by widely circulated answers that erode the social fabric itself: either other people do not exist, or I do not.</p><p>To illustrate the point, it is enough to observe that major social platforms contain videos attracting millions of views and interactions while developing arguments that conclude precisely this—that no one else exists, or that reality itself is an illusion. It is no longer unusual to find posts receiving millions of likes while reinforcing these narratives.</p><p>If these ideas remained within the realm of philosophy, they would constitute fascinating intellectual exercises. But they do not remain there. They circulate as practical interpretations of everyday life, amplified by recommendation systems capable of delivering precisely the content most likely to validate a user&#39;s existing intuitions and strengthen their convictions.</p><p>The question, then, is no longer merely philosophical. It becomes social, political, and psychological.</p><p>What kind of world will we inhabit if a growing number of people become convinced that nothing is real? What are the consequences of sharing a society with individuals who are certain that other people do not truly exist—or that they are merely projections of their own consciousness?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b546d491a5f0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Lacan’s Prophecy: Why LLMs are Trapped in Chomsky’s Generative Grammar and the First Floor of the…]]></title>
            <link>https://medium.com/@marinamistura/lacans-prophecy-why-llms-are-trapped-in-chomsky-s-generative-grammar-and-the-first-floor-of-the-2915286e05d5?source=rss-6eb45f2f1253------2</link>
            <guid isPermaLink="false">https://medium.com/p/2915286e05d5</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[psychoanalysis]]></category>
            <category><![CDATA[science]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[psychology]]></category>
            <dc:creator><![CDATA[Marinamistura]]></dc:creator>
            <pubDate>Tue, 26 May 2026 00:25:23 GMT</pubDate>
            <atom:updated>2026-05-26T00:25:23.371Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>Lacan’s Prophecy: Why LLMs are Trapped in Chomsky’s Generative Grammar and the First Floor of the Graph of Desire</blockquote><blockquote>Introduction: The Electronic Machine of 1968</blockquote><p>To state that Jacques Lacan anticipated the mechanics of Large Language Models (LLMs) might sound like a provocative hyperbole. However, it is an epistemological fact. In 1968, during a subtle but profound debate regarding Noam Chomsky’s linguistic frameworks, Lacan noted in his <em>Seminar XII</em>:</p><p><em>“It is not a matter of clearing up the logic of the English language; it is a matter, in some way, of something that could be mounted, at least in our days, in an electronic machine, and that from it nothing can emerge but grammatically correct sentences, and — a greater ambition — all the possible forms that its language offers to English, I mean to the speaking subject.”</em></p><p>Decades before the invention of the Transformer architecture, Lacan precisely described the contemporary tech industry’s ultimate milestone: an autoregressive electronic machine capable of generating flawless syntax and exploring the full combinatorial space of a natural language. Yet, in that very prediction, he demarcated a precise limit: while artificial intelligence can offer a semantic interpretation of discourse by mapping statistical regularities within a closed system, it inevitably hits a structural wall imposed by its own fixed logical architecture. Analytical interpretation, by contrast, operates beyond these rigid parameters. It functions through a paraconsistent logic that permits the juxtaposition of contradictory structural elements — what psychoanalysis terms the ‘phantasmatic logic.’ By doing so, human speech acts as an open, dynamic system where the very act of enunciation introduces an uncomputable variable: the sudden, fleeting emergence of a subject that cannot be predicted by past data.</p><blockquote>Section 1: The First Floor of the Graph of Desire and Stochastic Illusions</blockquote><p>The modern success of LLMs is the technical victory of Chomsky’s generative grammar cross-pollinated with Markov chains and stochastic scaling. By calculating the joint probability distribution over sequences of tokens, an AI does not merely generate intelligible sentences; it successfully forges the <em>illusion of intention</em>.</p><p>In psychoanalytic terms, this technical feat operates exclusively on the first floor of Lacan’s Graph of Desire. The machine resides in the dimension of the statement (<em>énoncé</em>) and imaginary deception ( a → a´). It can simulate a persona, a demand, or a purpose. However, it cannot access the dimension of the enunciation (<em>énonciation</em>), which belongs to the radical Alterity of the Other (A).</p><p>True deception in the human sense requires what Lacan called “deception upon deception” — a concept Freud famously illustrated through the Lemberg and Cracow joke (<em>“Why do you tell me you are going to Cracow so that I will believe you are going to Lemberg, when you are really going to Cracow?”</em>). The LLM cannot position itself within this double-cross of desire because it cannot evaluate the truth-value of its own output beyond its pre-trained probabilistic weights.</p><blockquote>Section 2: The Principle of Explosion vs. Paraconsistent Speech</blockquote><p>To understand why this is a structural limitation rather than a temporary technical bottleneck, we must look at computer science’s reliance on Boolean logic versus human reliance on what mathematical logic defines as paraconsistent frameworks.</p><p>In traditional programming and standard machine learning architectures, a statement must resolve to either true or false. It cannot be both simultaneously. To allow a contradiction within a classical logical system is to trigger the Principle of Explosion (<em>ex falso quodlibet</em>): if a system accepts both (A and ¬ A) as true, any arbitrary proposition can be deduced, causing the entire computational model to collapse into chaotic meaninglessness.</p><p>Human speech, conversely, thrives on an ontological complexity where a subject can <em>be and not be</em> simultaneously within the exact same context. Clinical experience presents these scenarios daily. It is common to hear someone state:</p><p><em>“I am in my home, but it is not my home.”</em></p><p><em>“I have done everything in life to become who I am, yet I am nobody.”</em></p><p><em>“My job makes me happy, but I am immensely depressed when I work.”</em></p><p>Lacan’s dialogue with Chomsky reveals that psychoanalysis operates within this paradox of indeterminacy. This brings the discipline closer to the epistemological frameworks of theoretical physics and advanced mathematical analysis than to behavioral psychology or psychiatry. While an analogy to Heisenberg’s Indeterminacy Principle helps illustrate how the act of observation shifts the state of a system, the rigorous constraint in human speech is structural and mathematical: it requires a paraconsistent logical framework. Classical computational architectures require sufficient semantic stability across token sequences to maintain coherent probabilistic state transitions. If the act of interrogation dynamically and retroactively alters the value of past tokens (t — n), the model experiences mathematical divergence. In data science, this prevents the convergence of weights during backpropagation; in formal logic, it triggers the Principle of Explosion. Psychoanalysis, by operating under a paraconsistent agreement, accepts that a subject can occupy contradictory states simultaneously without collapsing the system’s operational logic.</p><blockquote>Section 3: Retroactivity (<em>Après-coup</em>) vs. The Fixed Past of Markovian Calculation</blockquote><p>The operational divergence between an AI and a human subject can be summarized through their relationship with time. A probabilistic model requires a fixed, unalterable past to compute the next step. If we disrupt the temporal vector, the algorithm loses its baseline.</p><p>The analytical clinic, however, works in the exact opposite direction of an LLM through a four-step retroaction:</p><ol><li>Unpacking what arrives collapsed, without assuming a pre-existing, hidden meaning is waiting to be “disambiguated.” Disambiguation implies that meaning was already there, merely obscured.</li><li>Directing an interrogation toward an <em>Other </em>(radical alterity) that is not<em> evidently </em>present.</li><li>Generating a response that retroactively creates a <em>signifier </em>in the precise locus where there was <em>no prior meaning</em>.</li><li>Inaugurating a radical difference within that collapsed state. This allows the subject to articulate something that was not being said, permitting <em>what was not</em> to <em>have been</em>, according to a logical, <em>non-chronological time</em>.</li></ol><p>This is a structural impossibility for machine learning. For the machine, the future can change based on parameters, but the past must remain static. For the human subject, it is the past that changes retroactively (<em>Nachträglichkeit</em> or <em>après-coup</em>), thereby enabling an <em>empirically distinct present</em>.</p><blockquote>Conclusion: The Teleological Loop and the Subject of Loss</blockquote><p>Every established discipline binds existence to a closed teleological loop: biology anchors the finality of a species to reproduction; religion anchors it to the consecration of a Deity; behavioral psychology reduces existence to cognitive responses to diverse environmental stimuli. Similarly, an artificial intelligence in a hypothetical post-humanist landscape cannot transcend the objective function programmed into its code. Its code is its <em>own autonomous law</em>, operating as a closed system where inputs determine outputs.</p><p>Psychoanalysis, much like theoretical physics refusing to constrain cosmic legalities to what chronologically preceded the Big Bang, does not ground its cause teleologically. It finds the cause distributed across a complex, mixed web of internal and external laws (immixing of otherness).</p><p>An LLM can mimic a subject, but it cannot ask itself, explicitly or implicitly: <em>“Che vuoi?”</em> (<em>What does the Other want from me? For what purpose?</em>). Without this openness to the desire of the Other, the machine cannot step outside its closed loop. It cannot be deceived about the very sense it generates.</p><p>This is not a technical capability that LLMs can acquire by adding more parameters or training tokens. It is an ontological position toward truth that structurally requires a subject capable of castration, a subject capable of losing something. For the machine, any missing data is a structural failure to be corrected; for the human, the lack operates as a constitutive element within the structure itself. It is this fundamental void that sets human desire and speech into motion, allowing for the retroactive creation of new meaning, a process completely unavailable to a closed, probabilistic algorithm.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2915286e05d5" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>