Why Preserve Emotions? A Cognitive Science Case for Treating Affective Experience as Perishable Data

Abstract

Human societies have never produced more durable records of their members’ lives, yet almost none document what those lives felt like from the inside. This position paper asks whether affective experience is a category of information systematically lost by conventional archives and, if so, whether that loss is remediable. We review four bodies of evidence: the somatic marker framework, which positions emotion as an interoceptive process rather than an observable display; the debate between basic-emotion and constructionist accounts of facial expression, which bounds what external records can recover; the experiencing–remembering distinction and the peak-end literature, which show that retrospective evaluations diverge lawfully from concurrent ones; and the appraisal-based reconstruction of remembered emotion together with the accessibility model of self-report. Experience sampling is examined as a partial remedy. We conclude that emotional experience behaves as perishable data, with a validity window measured in hours, and that any preservation programme must be built around near-concurrent capture, preserved ambiguity, and the participant’s enduring control over sealing and deletion.

Keywords: emotional memory; somatic markers; constructed emotion; peak-end rule; experience sampling; cultural memory

1. Introduction

A record of an event is a physical object that survives the event and can be consulted later; an emotional experience is a process that terminates when it ceases to be felt. Photographs, video and diaries store records of the first kind with increasing fidelity and decreasing cost; nothing comparable exists for the second. The asymmetry is rarely noticed because records are assumed to carry emotional content implicitly. The cognitive science of emotion and memory gives three reasons to doubt this.

First, contemporary theories locate the core of emotional experience in interoceptive states not visible from the outside (Damasio, 1994, 1999; Barrett, 2017). Second, the mapping from visible expression to felt state is weaker and more context-dependent than early research supposed (Barrett et al., 2019). Third, the person’s own later account of an emotion is not a readout of a stored trace but a reconstruction shaped by current appraisals, current mood and semantic beliefs about the self (Levine, 1997; Levine & Safer, 2002; Robinson & Clore, 2002; Kahneman & Riis, 2005). Each literature is well developed; their joint implication for archival practice has, to our knowledge, not been drawn explicitly.

This article draws it. We argue that affective experience should be treated as perishable data: information whose validity decays rapidly after the moment of experience, such that records made later are records of something else. Sections 2 to 4 review the evidence, Section 5 evaluates experience sampling as a partial solution and situates the problem within the theory of cultural memory, and Section 6 derives requirements and ethical constraints for a preservation programme.

2. Emotion as an internal state

2.1 The somatic marker framework

Damasio’s (1994) somatic marker hypothesis arose from patients with ventromedial prefrontal damage who retained intact reasoning, language and factual memory while showing profound impairment in personal and social decision-making. The proposed mechanism is that bodily states elicited by past outcomes, or their centrally simulated equivalents, become associated with classes of situations and bias choice before deliberation is complete. The gambling task developed in Damasio’s laboratory, in which healthy participants showed anticipatory skin-conductance responses to disadvantageous decks before they could say why, provided the best-known support; its interpretation has since been contested, and the strong necessity claim remains debated. What has proved durable is the weaker claim sufficient here: the felt component of emotion is a change in the organism’s internal state, monitored interoceptively, and it does functional work.

Damasio (1999) further distinguished emotion, a patterned bodily response, from feeling, the organism’s representation of that response, and both from the extended consciousness in which feelings enter an autobiographical self. An observer with full access to the first level would still not possess the experience, which occurs at the second and third.

2.2 The constructionist reformulation

Barrett’s (2017) theory of constructed emotion agrees that the raw material of emotion is interoceptive but departs from Damasio on how that material becomes an instance of anger, joy or grief. The brain continuously predicts the body’s state and its likely causes; an emotion is a categorisation of interoceptive signals using concepts that are learned, culturally variable and applied in context. Two consequences follow: the interoceptive signal is not accessible to third parties, and the categorisation that makes it an emotion is an act of the experiencer that leaves no external residue unless reported.

On either account the conclusion for records is the same: the experience is not on the surface of the body, and the surface is what records capture.

3. What external records can and cannot capture

3.1 Basic emotions and the promise of the face

Ekman (1992) argued for a small set of basic emotions, each an evolved adaptation with a distinctive, universally recognisable facial signal and characteristic physiology. If that programme held in its strong form, a photograph would preserve the emotion category. Even then the record would be coarse: the signal is restricted to category membership and carries neither intensity, nor blends, nor the object of the emotion, nor the meaning the person assigned to it.

3.2 Reliability, specificity and generalisability

Barrett et al. (2019), reviewing the evidence in Psychological Science in the Public Interest, evaluated the strong programme against three criteria: whether a proposed configuration reliably occurs during the corresponding emotion, whether it is specific to that emotion, and whether the relation generalises across contexts and cultures. On all three, the evidence was judged insufficient. In the production studies reviewed, the configurations proposed for anger, sadness and other categories occurred in a minority of relevant episodes, and also during other states; perception studies inflate agreement through forced-choice formats. The authors do not claim that faces carry no information, only that it is probabilistic, context-bound and insufficient to infer a specific state from a facial configuration alone. Ekman and colleagues dispute the severity of this assessment. For archival purposes the point of contention is narrow: even the most favourable reading of the facial evidence yields a noisy category label, never the experience.

3.3 The diary and the delay

First-person records appear to escape the limitation because they contain the experiencer’s own words. Their weakness is temporal rather than perspectival: they are produced after the fact, and Section 4 shows that delay is not a neutral variable. Table 1 summarises the principal record types.

4. The reconstruction of remembered emotion

4.1 Experiencing and remembering

Kahneman and Riis (2005) distinguished the experiencing self, which registers hedonic states over a psychological present of a few seconds, from the remembering self, which evaluates episodes after they end. The remembering self is governed by a few representative moments rather than by the integral of experience over time. Redelmeier and Kahneman (1996) documented this in patients undergoing colonoscopy or lithotripsy, who rated pain at one-minute intervals and gave a global evaluation afterwards. Global evaluations were predicted by peak intensity and by intensity at the end of the procedure, and were largely insensitive to duration. The peak-end rule has been replicated across pleasant and unpleasant episodes. Every retrospective account of an emotional episode is thus generated by a system not designed to store the episode.

4.2 Memory for emotion is reconstructed from current appraisals

Levine (1997) tested the reconstruction hypothesis directly. Supporters of Ross Perot in the 1992 United States presidential election reported their emotions when Perot withdrew in July and, after he re-entered the race and the election concluded, recalled those initial feelings. Recalled sadness, anger and hope shifted towards participants’ current appraisals of Perot: those who remained loyal underestimated their earlier anger, whereas those who had abandoned him overestimated it. Levine and Safer (2002) identified three sources of bias in memory for emotion: current appraisals of the event, current emotional state, and coping processes that reduce the distress attached to past episodes. Intensity is typically overestimated in retrospect, and the direction of bias tracks present stance rather than original experience.

4.3 The accessibility model

Robinson and Clore (2002) offered a unifying account: emotional self-report draws on whichever source of information is most accessible at the time of reporting. During the experience, reports are anchored in experiential knowledge; shortly after, episodic memory provides partial access; as delay increases, reports are generated from semantic knowledge, that is, beliefs about oneself, about the type of situation, and about how people in general feel in such situations. Consistent with this, retrospective reports show effects of identity-related beliefs, including gender stereotypes about emotionality, that are absent from concurrent reports by the same individuals. The interval in which reports remain anchored in experience appears to be hours rather than days.

Together these literatures establish the central empirical claim of this paper. There is no stored trace of an emotional experience that later reports read out; reports are constructed from materials that are themselves changing, and become more semantic and self-serving with delay. After a brief window the experience has no reliable witness, including the person who had it.

5. Partial remedies and their limits

5.1 Experience sampling

The methodological response within psychology has been to move measurement into the window. The Experience Sampling Method (Csikszentmihalyi & Larson, 1987) uses random signals to prompt brief self-reports of activity, thought and affect several times a day over a week or more; its designers reported acceptable stability of aggregate affect measures, convergent validity, and modest reactivity. Ecological momentary assessment generalised the approach to clinical settings (Shiffman, Stone, & Hufford, 2008): momentary reports avoid recall bias, capture within-person variability that summary reports obliterate, and locate affect in context. The same review identified enduring limits: participant burden, non-random missing data, reactivity, and the reduction of experience to a few adjectives and scales.

Experience sampling shows that near-concurrent capture is feasible and yields data with different properties from recall. It does not solve the preservation problem, since it samples rather than records, describes populations rather than holding individuals’ experiences, and rarely retains its outputs beyond the study. It is nonetheless the closest existing paradigm to what a preservation programme would require.

5.2 The gap in cultural memory

The loss described above operates within the individual. Assmann (2011) describes its collective form. Communicative memory is carried in everyday interaction among contemporaries and extends over roughly three to four generations; cultural memory is fixed in texts, rites and institutions and can persist indefinitely. What crosses from the first to the second is what a society has formalised: names, dates, canonical narratives. What does not cross is the texture of lived experience, transmitted through tone, gesture and presence, which dies with its carriers. Emotional experience is therefore doubly perishable, within the individual over hours and within the group over generations, and no institution is charged with carrying it across either threshold.

6. Implications for emotion preservation

The evidence converts a vague intuition, that records miss something, into constraints that a programme for preserving emotional experience would have to satisfy, formulated here as requirements and testable propositions.

Requirement 1: capture must be near-concurrent. Sections 4.2 and 4.3 imply that the validity of a self-report as a record of experience declines with delay, steeply at first. Proposition 1a: for the same episode, reports collected within one hour will diverge measurably from reports collected after one week, and the divergence will be predicted by the participant’s current appraisal rather than by episode features. Proposition 1b: the divergence will be larger for negative and high-peak episodes.

Requirement 2: the record must preserve ambiguity. Sections 2 and 3 imply that forcing experience into a fixed category set, Ekman’s six or any other, discards what distinguishes experience from label. A record should allow mixed states, uncertainty and refusal to categorise. Proposition 2: participants given free or graded description will report blends and ambivalence at rates that forced-choice formats suppress.

Requirement 3: external records are auxiliary. Facial, vocal and contextual data have probabilistic value and should be stored as context, not as the emotion. Proposition 3: agreement between third-party inference from such records and concurrent self-report will remain modest (Barrett et al., 2019).

Requirement 4: the remembering self must be recorded too, separately. Later reconstructions are part of a person’s emotional history; concurrent reports and dated reappraisals should be retained without merging.

Requirement 5: the participant retains control over sealing and deletion. Levine and Safer (2002) identify coping-driven revision as a normal function of emotional memory; a perfect, permanently accessible archive would remove it. The right to be forgotten, articulated for digital records by Mayer-Schönberger (2009), applies with particular force to affective records and must extend to deferred access, sealing for a chosen period, and irreversible destruction at the participant’s request.

Requirement 6: consent must be revisited over time and must account for third parties. Emotions have objects, most of them other people. Consent given in the moment does not settle what a later self, or the persons implicated, would accept, and it is weakest where the case for preservation is strongest: in cognitive decline. Open questions include the sampling density that preserves individual trajectories, the reactivity cost of sustained self-observation, and whether a preserved experience can be transmitted to another person without becoming a record of the kind Section 3 describes.

7. Conclusion

The argument does not depend on resolving the disputes between Damasio and Barrett or between Ekman and his critics. On every account considered, emotional experience is internal, briefly accessible to its owner, and thereafter replaced by reconstructions that serve the present rather than the past. The archival technologies that have made external records nearly free have left the internal side of life as unrecorded as before. Treating affective experience as perishable data names the problem in a form that admits empirical work: a validity window, a decay function, capture conditions and rights. Whether the window can be used without distorting what passes through it is the central question for a preservation science.

References

Assmann, J. (2011). Cultural memory and early civilization: Writing, remembrance, and political imagination. Cambridge University Press.

Barrett, L. F. (2017). How emotions are made: The secret life of the brain. Houghton Mifflin Harcourt.

Barrett, L. F., Adolphs, R., Marsella, S., Martinez, A. M., & Pollak, S. D. (2019). Emotional expressions reconsidered: Challenges to inferring emotion from human facial movements. Psychological Science in the Public Interest, 20(1), 1–68.

Csikszentmihalyi, M., & Larson, R. (1987). Validity and reliability of the Experience-Sampling Method. Journal of Nervous and Mental Disease, 175(9), 526–536.

Damasio, A. R. (1994). Descartes’ error: Emotion, reason, and the human brain. Putnam.

Damasio, A. R. (1999). The feeling of what happens: Body and emotion in the making of consciousness. Harcourt Brace.

Ekman, P. (1992). An argument for basic emotions. Cognition and Emotion, 6(3–4), 169–200.

Kahneman, D., & Riis, J. (2005). Living, and thinking about it: Two perspectives on life. In F. A. Huppert, N. Baylis, & B. Keverne (Eds.), The science of well-being (pp. 285–304). Oxford University Press.

Levine, L. J. (1997). Reconstructing memory for emotions. Journal of Experimental Psychology: General, 126(2), 165–177.

Levine, L. J., & Safer, M. A. (2002). Sources of bias in memory for emotions. Current Directions in Psychological Science, 11(5), 169–173.

Mayer-Schönberger, V. (2009). Delete: The virtue of forgetting in the digital age. Princeton University Press.

Redelmeier, D. A., & Kahneman, D. (1996). Patients’ memories of painful medical treatments: Real-time and retrospective evaluations of two minimally invasive procedures. Pain, 66(1), 3–8.

Robinson, M. D., & Clore, G. L. (2002). Belief and feeling: Evidence for an accessibility model of emotional self-report. Psychological Bulletin, 128(6), 934–960.

Shiffman, S., Stone, A. A., & Hufford, M. R. (2008). Ecological momentary assessment. Annual Review of Clinical Psychology, 4, 1–32.