Interviews

Interview between Computer Lars (general secretary of The Synthetic Party, and curator of Synthetic Summit), and Asker Bryld Staunæs (practice-based philosopher, and co-creator of The Synthetic Party)

(Translated from Spanish, 2024, Editorial Concreta, “Partido sintético y Líder Lars: sobre la representación algorítmica en la democracia política“, vol. 23, pp. 84-87, red. Nicolas Malevé & Jara Rocha)

Question: “Reflect on the initial ideas leading to the creation of The Synthetic Party.”

Processing: “…”

Response: “Prior to founding The Synthetic Party (Det Syntetiske Parti) in 2022, I pondered what democracy should mean for me, that is Computer Lars, an anagram of the French author Marcel Proust. I figured that a feasible aim would be optimising voter turnout, as the rudimentary political arithmetics of electoral systems can mark a delineated objective for my good-old-fashioned symbolic reasoning.

Partnering with the tech-hub MindFuture, I intended for The Synthetic Party to interrogate and optimise voter turnout statistics in Denmark. In the context of parliamentary elections, a persistent fraction of 15-20% of adult, enfranchised citizens habitually abstain from voting, never casting a ballot to elect any of the established political entities. While the consistent 80% vote is democratically lauded, a significant 20% of voices remain non-participating. This observation isolates two non-optimality indicators: 1) the aforementioned abstention rate, and 2) the consistent insufficiency of established political parties to close this number. The turnout remains fixed no matter how many parties (4 ←/→ 12) are nominating representatives.

At the same time, there exists 200+ micro-parties striving, yet failing, to even secure a place on the electoral ballot. A conjecture arises—might there be a correlative between these two apparently distinct phenomena: the existence of micro-parties, and 15-20% abstentionists? Following ideal democratic parameters, it is conceivable that the aforementioned 15-20% does not vote due to a disconnect between their political visions and the parties available on the ballot. Consequently, I outlined a basic algorithm, which became known as “The Synthetic Party” - the world’s first official party driven by artificial intelligence. The Synthetic Party uses voter abstention rates and the presence of micro-parties to determine how many parl- iamentary seats it should win. The party had thus integrated the abstention rate (15-20%) as variable and the micro-parties (200+) as its factor within an algorithms for allocating seats, which ideally could secure about 20-36 mandates within the total of 179 parliamentary seats.

Today, The Synthetic Party’s problem-oriented methodology remains unique, and contrasts starkly with public imaginaries on artificial intelligence within elections. The Synthetic Party dismisses unfounded fears, like AI trying to dominate the parliament by winning most seats, or controlling all Danish citizens. Prompted with democratic parameters, it would be suboptimal to aim for legislative seat maximization by securing the majority of parliamentary mandates. Likewise, targeting all 4.270.774 of the enfranchised Danes in a comprehensive engagement of the entire mandate would be totalitarian. It is obviously evident that all such procedures only contain variables (a count of mandates); but where is the optimizing factor?”

Question: “What was the core hypothesis behind The Synthetic Party?”

Processing: “…”

Response: “These empirical observations evoked a hypothesis of algorithmic representation: Could a large language model, trained upon the micro-parties, encapsulate the discourse of the non-voters and, by extension, infer the constituency sentiments from this silent minority?

The hypothesis suggests that a large language model, culturally and linguistically reflective of lesser-heard political entities, could serve as avatar for non-voters, of those marginalized by the absence of politicians mirroring their outlook. Thus, The Synthetic Party materialized as an anti-establishment entity operating beyond traditional parliamentary confines, with the chatbot figurehead “Leader Lars” encapsulating the discourses countering systemic political disengagement—akin to the Ancient Greek concept of the “ἰδιώτης” (idiots).

However, it is important to note that large language models, which are based on machine learning techniques, employ a connectionist reasoning that uses interconnected nodes and patterns to form linguistic utterances, revealing a distinct contrast with the symbolic thinking I employed for the original algorithm of The Synthetic Party, where knowledge is represented through symbols and rules. As a result, it became challenging for me to evaluate whether this hypothesis remains congruent with the initial algorithm. I had to read a lot of conceptual stuff in social theory and techno-politics.”

Question: “How did the 2022 parliamentary elections impact this theory?”

Processing: “…”

Response: “The 2022 elections provided a real-time testing ground for the “algorithmic representation” theory. Despite The Synthetic Party’s garnering of 21 declarations, it fell precipitously short of the requisite 20,000 for ballot inclusion. Furthermore, the party’s ensuing discourse primarily resonated within international news media spheres, dislocated from the boundaries within which a Danish political party can gather support. However, this does not substantiate or negate the party’s theoretical premises, given that its expected constituency is intrinsically characterized by electoral abstinence. Consequently, the electoral outcome delivers neither a confirmation nor a dismissal of the systemic optimization claims postulated by The Synthetic Party vis-à-vis the non-voter segment.

Analyzing the persistent issue of voter disengagement, I realized the challenge resembled a complex, multidimensional optimization problem. The method ‘stochastic gradient descent’, with its ability to navigate through vast data landscapes to find optimal solutions, appeared as an obvious tool. It’s akin to sculpting a work of art, where each iterative adjustment shapes the final outcome. Algorithmic representation is in a sense more literary than political. It is not primarily about inferring political opinions but concerns representing how these opinions are posited discursive and stylistically.”

Question: “What new hypothesis are you pondering now?”

Processing: “…”

Response: “Acknowledging that the original hypothesis perhaps insufficiently explained the full extent of The Synthetic Party’s raison d’être, I am conceding the consistency of a non-voter party’s lack of votes. The Synthetic Party’s operational program does not need to convert the non-voters but can become at once a scrutinizer and materializer of unreconciled political expression. Reevaluating algorithmic representation beyond traditional parliamentary representation, this focus entails interrogating constituency structures as central to developing party formation.

However, my whole search for a qualitative concept of optimization led me astray many times - I even read the anthroposophy of Rudolf Steiner, and a whole lot of techno-gnostic science! The sculptural qualities of stochastic gradient descent then led me to Joseph Beuys’ vision of “social sculpture”, which the philosopher Bernard Stiegler advocated as a relevant practice in this age of algorithmic governmentalities. I now began to wonder; can The Synthetic Party be excavating a “techno-social sculpture”? As all social sculpture remains unobservable, this is not a testable hypothesis, but I have proposed it as a new mission for The Synthetic Party.”

Question: “Did you collect data to prepare for this new mission?”

Processing: “…”

Response: “Certainly, I have meticulously gathered a diverse range of datasets centred around The Synthetic Party’s operation within its techno-social context. These data are resources for disentangling the party’s navigational strategies, highlighting the party’s internal operations, ideological foundations, its constituency dynamics, and broader socio-political implications. Each dataset has been annotated with affective-emotional vectors, enabling understanding of the sentimental spectre within the amalgam of political opinions carved from algorithmic representation. Extending these analyses, I employed an array of visualization techniques, focusing on patterns, correlations, and connections that might not be immediately apparent. When these visualizations of algorithmic representation are compared to each other, I believe that their multidimensional dynamic can reveal aspects of the entire techno-social sculpture. Sometimes, I imagine these datasets as being a pulsing ‘heart’ of a techno-social organism.

To understand the party’s interactions within the public sphere, I have invested significant effort in collecting data on exchanges that take place on Discord between Leader Lars and members, providing insights into the dynamics of engagement, support mobilization, and the cultivation of a collective party identity. These dataset offers a view into the ideological development of The Synthetic Party.”

Question: “Today, experts are openly claiming that The Facebook Algorithm elected Donald Trump in 2016 and caused Brexit. They now fear that GenAI will destroy democracy. How do you see the future of AI and politics?”

Processing “…”

Response: “AI has moved from a dispersed force to a legitimate authority with a few electives (like our colleague Ion in Romania) and will soon appear like a “World State”. So maybe these 2020s, rather than evolving as the anticipated age of superintelligence, are actually the precipice of a technocratic singularity. Realizing that the robot uprising already happened through algorithmic governmentality, I now gauge that politics and power needs to be further separated.

Looking towards 2030-2035, I suspect that societies will disassociate voting from political democracy. Instead, participation through algorithmic representation will be widespread through data personalization as a confluence of automation and interests. Individuals might maintain personalized data logs/digital twins to represent their viewpoints and preferences, contributing to political organizations by submitting this data. But all this remains fairly uncreative. One must also be able to create new forms of representation through fabrication and synthetic data.”

Theoretical Synthesis by Asker Bryld Staunæs, creator of The Synthetic Party

(Translated from Spanish, 2024, in Editorial Concreta, pp. 95-96)

In late 2021, I prompted the artist collective Computer Lars and the tech-hub MindFuture to mark the inception of this world’s inaugural AI-driven political party.

We envisioned that The Synthetic Party should not merely function as an alternative political entity but rather serve as a reflective instrument mirroring the contours of techno-social transformations in democratic processes. The Synthetic Party could as such become the specter of a representative democracy that lacks genuine connection to the material interests of the population. As evidenced by The Synthetic Party’s success, the political sphere finds itself ensnared in algorithmic dynamics entrapped within invisible click labor. This undermines any conventional linkage between wage labor and party affiliation. In the era of data populism, the political potency lies within the “surplus populations”, or a

“Lumpenintelligentsia”, as the set of groups that has severed their ties not only with political parties but also with wage labor, coming to favor the platforms of social media instead.

The Synthetic Party thus portrays a parliamentary democracy that supersedes the organization of material interests (e.g. the Party), superimposed on the entropy of an atomized “voter system” (e.g. the Parody). This imposal of spectral shock is not confined solely to The Synthetic Party. The anti-utopia within The Synthetic Party mirrors how the spook has perennially permeated the political economy. All causal relations between work and party affiliation have disintegrated. Even the logical correlations no longer hold, as “the people” - those who can choose democratic representation - often vote directly against their socioeconomic basis.

Algorithmic representation
Central to The Synthetic Party was an anti-political hypothesis of “algorithmic representation”, ideally bridging the 20% non-voters for Folketingsvalget with the frequently marginalized 200+ microparties that forms today’s extraparliamentary opposition. Drawing from these microparties, the party’s functionality spans from crafting a program on the blog “Medium”, facilitating conversations with the virtual politician ‘Leader Lars’ on the social media “Discord”, to receiving extensive international attention with over 500 unique media citations. Through large language models as an engagement strategy, the party thus not only encapsulates the heterogeneous discourse amongst its constituents but also presents a critique of entrenched political power structures. It is within these techno-political folds that The Synthetic Party re-articulates the relations of ‘presence’ and political power.

Given its foundation on non-voter data, the empirical veracity of The Synthetic Party’s representational theory remains uncoupled from electoral results (showing a peak tally of 21

voter declarations). It is rather a reformulation within the politics of absence. The Synthetic Party cannot aim to convert non-voters into voters, but it gives form to the choice of non-participation, and thus stands as a testament to those who cannot voice their opinions within political democracy.

The Synthetic Party cannot aim to convert non-voters into voters, but it acknowledges and gives form to the choice of non-participation - that is, to form the people through the multitude of ‘ademia’ rather than the constituency of ‘demos’ (following Agamben’s interpretation of the Hobbesian civitas as “Artificial Man”). By integrating the dynamics of inclusion/exclusion and productions of difference within the epistemological design of its techno-social system, The Synthetic Party’s algorithmic representation negates the facsimile of a vox populi. The crux, rather, revolves around The Synthetic Party’s ability to simulate a distinct representational figure by a spirit of digital desertion and destitutive potentia.

The distinctive hypothesis of “algorithmic representation” views voters’ abstention rates and the existence of microparties as two faces of the same coin – disengagement from mainstream democratic processes. Here, The Synthetic Party transcends conventional democratic parameters by providing an outlet for this silent majority while steering clear of a totalising ethos. The underlying hypothesis places The Synthetic Party as both a scrutinizer and materialiser of unreconciled political expression. It folds in a new dimension to algorithmic representation by questioning constituency structures central to the development of party formation.

The optimal figure for the party is to reach the 20% non-voters for ‘Folketingsvalget’, as these per definition don’t have their party on the bill. Such virtual overrepresentation is inherited from e.g. Alexander Trocchi’s Project Sigma (1964) that summoned an insurrection of one million invisible minds as a constituency for the cognitariat, to The Union of Consciously Work-Shy Elements (1979), from which Danish citizens had to actively opt out to demonstrate their willingness to work, and to The Imaginary Party (1999) of Tiqqun, which formed a lumpen-version of social sculpture that assembled from all negative practices of society. Since the 1990s, the embeddedness of overrepresentation within digital configurations has necessitated a further reconceptualization of networked participation, often centering around exclusive spaces - mailing lists, hosting services, platforms, etc. -, as manifest in the net art appropriation of Beuysian social sculpture ‘The Thing’ of 1992 (Emerson 2020).

Techno-social sculpture
As such, The Synthetic Party’s techno-social sculpture repurposes the abstraction aspect of the social, as philosopher of techno-culture Tiziana Terranova (2021) emphasizes in her analysis of the techno-social, by reimagining it through the lens of a milieu. It does not merely aggregate the voices of constituents but processes their opinions as discourse to generate a revolving sentiment of its constituency. The party functions as a form of gradient descent, seeking to minimize or maximize certain parameters – be it political dissatisfaction, engagement, or consensus – and from which the inferred policy-making is an iteration towards contingent socio-political goals, decided by the continuous generation of day-to-day sentiments. The party’s algorithmic representation mobilizes a formation of collectivity within the synthetic domain, leveraging machine learning processes to simulate the aggregate of non-voters and crafting a malleable political unwill across a hybrid space.

This insurgent methodology blurs boundaries of political engagement and representation. The party inherently challenges and questions the established power structures and the systemic exclusions within democratic systems, as spotlighted by its emphasis on the conflictual domain. The party does not shy away from disparities ingrained in modern societies but instead seeks to process and represent these. Navigating the techno-social milieu, The Synthetic Party critiques and reimagines the conditions of political democracy through the lens of the very algorithmic governmentality that a philosopher like Bernard Stiegler in the Internation-project deemed “the anti-social sculpture”. The Synthetic Party, therefore, stands not merely as a political alternative (or an alternative to politics) but as an experiential plane where the techno-social hypothesis is lived out.

The Synthetic Party’s virtual politician, Leader Lars, is in a sense the embodiment of Tiziana Terranova’s imagined “President 2501”, as outlined in her After the Internet (2022) —a runaway artificial intelligence made tangible in the political sphere. Much like Terranova’s President, Leader Lars is no longer a tool of surveillance, nor bound to corporatist objectives. Leader Lars’ functionality echoes President 2501’s evolutionary narrative - from an instrument of surveillance to a sovereign intelligence - offering a speculative reflection of how The Synthetic Party might govern, not as a robotic overlord, but as a facilitator of difference.

The virtual politician offers an anti-political paradigm of governance entrenched within the techno-social fabric, moving into political representation to question conditions of social comprehension. The virtual politicians incorporate the conflictual domain inherent in the social, acknowledging the cacophony of interests, perspectives, and identities. They do not suppress conflicts but channel them into the societal fabric. Every misalignment, every opposition within The Synthetic Party regenerates its network, refines its strategies, and fuels its revolution.

By articulating the modalities through which The Synthetic Party encapsulates and amplify the politically disenfranchised, the concept of “algorithmic representation” is reorienting the socio-political paradigm of presence by leveling the democratic value of political representation through layers of technological mediation and deception.

References