Our target platform was finally nailed down, but we had no time to rest. This is because we knew that the biggest decision of the entire project stood before us: the selection of the full pop song for inclusion in our intro.
We recognised that a perfect song was required for our audience. However, we also knew that this audience was not a unified one. Despite its apparently apolitical stance, we understood the demoscene to be split into two rival, and highly polarised, counter-factions.
At moments like this, science and logic must prevail over panic and fluster. To ground ourselves in the facts, we adopted a behavioural science approach to develop "pen-profiles" of each faction:
-- Type A: The Moral Majoritarian - Individuals aligned with this personality type adopt online personas characterised by progressive attitudes towards demoscene diversity, alongside high levels of group sensitivity towards opposing views. On detection of dissent, type As behave "as-one" in waves of swarm-like messageboard attacks, thereby disorienting, overwhelming and ultimately isolating their opponents.
-- Type B: The Lone Wolf Contrarian Edgelord - Unlike the more common type A, type Bs operate alone and are exclusively male. They will have flirted with at least two of the following ideologies: digital libertarianism, Men's Rights Activism, esoteric nationalist politics, and/or wildly unhinged conspiracy theorising. The online modus operandi of the type B is shitposting, be it to defend real-life acts of discrimination, escalate "anti-woke" dog-whistles, or argue that night is day.
The objective that emerged was complex. Somehow, we had to choose a full pop song which would not invoke adverse reactions from either personality type, instead cleanly sailing through the grey centre-ground of public opinion. This would protect us from the worrying risk of unpleasant or, even worse, potentially upsetting internet arguments.
To achieve our objective, we undertook a speculative root cause analysis. Here, we "war-gamed" type A and B responses to randomly-selected full pop songs via an iterative simulation technique. By undertaking ~450 rounds of simulation, we isolated key factors which loaded to the highest probability of positive audience response, regardless of group constitution.
![]() |
| Figure 1: Cumulative output of war-gaming simulation |
1) Songs with no hint of generative AI whatsoever (Cohen's d = 0.79, p<0.005). For safety, we therefore decided to exclude any song written since 2010 from our long list.
2) Songs containing no NSFW content (Cohen's d = 0.64, p<0.005). Across every simulation, at least one collective of types A and B was viscerally offended by NSFW material in any given full pop song, albeit for entirely different ideological reasons.
3) Apolitical song themes (Cohen's d = 0.623, p<0.005). Across all simulations, the exclusion of full pop songs expressing any political stance (even if unconsciously) maintained an attractive pretence of political neutrality for the audience.
4) Songs that fuse innovative musical genres (Cohen's d = 0.856, p<0.005). Most strikingly, all simulations demonstrated that maximum agreeability across both type A and type B personality types was achieved by full pop song stimuli that "crossed" (or "mashed-up") otherwise niche musical genres.
By now, the intellectual complexity of choosing the perfect full pop song for our intro was beginning to slip from our grasp. There were now simply too many moving parts for the human brain to comprehend. With much regret, we therefore took the potentially controversial decision to outsource the full pop song shortlisting to a generative AI model.
We did not take this decision lightly. Thankfully though, by pooling our collective expertise we engineered a series of replicable prompts, and merely a day later we were sat before a shortlist of four full pop songs. This is summarised in the table below, alongside our hand appraisal of each candidate song against our behaviourally-derived criteria:


No comments:
Post a Comment