Updating evaluations through memory editing approaches during wakefulness and sleep

Abstract

Human evaluations, such as likes and dislikes, significantly influence our decisions and behaviors. Evaluations are often shaped by memories of learning experiences, either from direct personal experiences (non-social evaluative learning) or from observing others (social learning). In this ever-changing world, evaluative information can be outdated or invalid, emphasizing the necessity for evaluation updating. This thesis investigates how evaluations can be updated through memory editing approaches during both wakefulness and sleep stages and in both social and non-social learning contexts.

Chapter 2 examined how social learning from ingroup and outgroup members affected the evaluation updating of facial attractiveness. Results showed that both types of learning updated attractive ratings; however, only learning from ingroup members updated neural representations of facial attractiveness. Moreover, evaluation updating was tightly associated with participants’ memories of the social learning experience.

Chapter 3 employed targeted memory reactivation (TMR) during non-rapid-eye-movement (NREM) sleep to update evaluations via reactivating memories of social learning. Results indicated that TMR induced overnight evaluation updating, mediated by cue-elicited delta-theta power (1-8 Hz) and overnight N2 spindle density. This study suggested that sleep-mediated memory reactivation processes drove this effect.

Subsequent studies employed a non-social evaluative learning paradigm, in which participants learned evaluative associations between hypothetical pharmaceutical products and positive or negative health outcomes.

Chapter 4 studied the impact of retrieval suppression on the evaluations of pharmaceutical products that were initially paired with either positive or negative health outcomes. The results demonstrated that retrieval suppression rendered these products less preferred, i.e., suppression-induced devaluation, regardless of the initial valence. This devaluation was accompanied by decreased evidence accumulation speed and prolonged non-decision times, as evidenced by the drift-diffusion model (DDM). The findings suggested that retrieval suppression resulted in devaluations in a valence-independent manner.

Chapter 5 investigated the impact of counter-attitudinal learning and TMR on evaluation updating. Results demonstrated that cueing counter-attitudinal learning reduced ambivalence levels during evaluation, as evidenced by mouse-tracking measurements, and facilitated evaluative memory integration when positive information was used to counter initially negative learning. Moreover, cue-elicited delta power (1-4 Hz) and spindle activity during sleep were associated with reduced evaluation ambivalence and enhanced integration of evaluative memory, respectively.

Collating findings from these studies, this thesis proposes a memory-based evaluation framework in Chapter 6. Specifically, evaluative decisions can be guided by mnemonic processes, including retrieving episodic memories (positive or negative outcomes) and ease of retrieval. Memory editing approaches, such as retrieval suppression during wakefulness and TMR during sleep, can impact memory processing (e.g., enhance, weaken, integration), thereby updating evaluation. Moreover, social, affective, and motivational factors can interact with the mnemonic processes to bias evaluation formation and updating. Based on the abovementioned findings and the proposed framework, potential memory-centered interventions can be designed to address decision-making deficits in clinical and healthy populations, such as social avoidance in social anxiety patients and substance abuse in addictive disorders patients, among others.