LINER NOTES · 2026-09-12

Music, preference & the person listening.

Research-informed listening, with a deliberately playful 16-type presentation. A portrait of this session’s choices—not a clinical test or an official MBTI assessment.

What the research supports

Rentfrow and Gosling found broad preference patterns and associations with personality [1]. Later work proposed the five-factor MUSIC model using musical excerpts: Mellow, Unpretentious, Sophisticated, Intense and Contemporary [2]. Greenberg and colleagues organized perceived musical attributes into arousal, valence and depth [3]. These are different models, not independent validations of MusicBite.

Our design inference: let people respond to sound itself, not only genre labels; describe the resulting preferences along intelligible dimensions. The papers below are conceptual references, not evidence that our items, generated clips, weights or cutoffs are validated.

Four listening dimensions

From choices to 16 personas

1. The live pool has 32 questions. A session selects 20 distinct questions with distinct presentation types. The first 10 are balanced across chapters and axis coverage; results are available after 10 completed questions. The extension adds the remaining presentation types, rather than repeating the first-half balancing constraints.
2. Every legal choice has a four-number score vector assigned by the product design. In a two-track duel, a mild preference contributes half the clip’s vector and a strong preference contributes the full vector. Other modes use their assigned option vector. Animations, gesture skill and time spent playing are not scored.
3. We sum each axis over the questions included in the result. For display, divide that sum by the sum of each included question’s maximum absolute possible contribution on that axis. This is a relative preference score, not a percentile or confidence probability.
4. A positive total selects K, L, A or X; a negative total selects S, D, I or H. An exact tie uses the first nonzero contribution in session order; an axis with no contribution defaults to its negative letter. Ties are therefore order-sensitive, not evidence of a stable type.
5. Four binary labels yield 2⁴ = 16 combinations. Names such as Engineer and Sampler are editorial metaphors, not empirically discovered personality clusters or career advice. The “neighbor” flips the axis closest to zero; it is not a statistical confidence interval.

Example: a positive first-axis score, negative second-axis score, positive third-axis score and negative fourth-axis score produce KDAH. This code looks up a name; the name does not determine the scores.

Why this format—and what it cannot claim

The four-axis format makes a continuous preference profile easy to explain and share. “Music MBTI” describes that interface convention only. We do not administer or reproduce MBTI, STOMP or the MUSIC inventory. We have not established test–retest reliability, factor structure, normative percentiles, predictive validity or cross-language measurement equivalence for MusicBite. Twenty answers provide more observations, but do not by themselves prove greater accuracy.

Familiarity, mood, playback equipment, culture and properties that vary together in a clip may influence choices. Generated excerpts are not tightly controlled laboratory stimuli. The same person may receive a different type on another visit. Artist associations and recommended songs are curated interpretations—not assessments taken by those musicians or scientific evidence for a type.

Friend links & listening matches

Sharing adds a public mix configuration to the URL. It contains four normalized scores rounded to 1% increments, a format version and an error-detection checksum—not a name, account ID or individual answers. It is readable and editable by anyone with the link, not encrypted or authenticated. Identical profiles can produce identical links.

Matching compares the four normalized scores with equal weight: 100 × (1 − sum of absolute axis differences / 8), rounded to a whole number. Each axis ranges from −1 to +1. The percentage is a playful similarity index, not a relationship prediction. Captions use bands of 85+, 65–84, 45–64, 25–44 and below 25; the 85+ band gets a party caption when both energy scores exceed 0.35. Friend data never changes your answers or personality score.

好友链接公开携带四维归一化分数，不含姓名和逐题答案；短码不是身份验证，同样的分数可得到同样的链接。匹配度＝100 ×（1 − 四轴绝对差之和 ÷ 8），取整数。它只是等权听感相似度，不预测关系好坏。“互补”“打起来”等都是按分数区间分配的趣味文案。好友数据不影响你的测试计分。

Music & provenance

The current pool contains 83 short audio clips. Production notes describe Google Lyria 3.5 generation from chord, style and contrast prompts, followed by excerpt selection and loudness/fade processing. These are not commercial recordings or a song-recognition test. Shared prompts do not guarantee matched melodies, harmony or all other acoustic properties.

Read the music production and usage notes. Public playback does not grant a license to redistribute the assets. Sponsorship and recommendations do not alter score calculation.

中文 · 理论基础与设计依据

16 names · 16 个名字

References / 参考文献

1. Rentfrow, P. J., & Gosling, S. D. (2003). The do re mi’s of everyday life: The structure and personality correlates of music preferences. Journal of Personality and Social Psychology, 84(6), 1236–1256. DOI: 10.1037/0022-3514.84.6.1236
2. Rentfrow, P. J., Goldberg, L. R., & Levitin, D. J. (2011). The structure of musical preferences: A five-factor model. Journal of Personality and Social Psychology, 100(6), 1139–1157. DOI: 10.1037/a0022406
3. Greenberg, D. M., Kosinski, M., Stillwell, D. J., Monteiro, B. L., Levitin, D. J., & Rentfrow, P. J. (2016). The song is you: Preferences for musical attribute dimensions reflect personality. Social Psychological and Personality Science, 7(6), 597–605. DOI: 10.1177/1948550616641473