Intermediate-term emotional bookkeeping is necessary for long-term reciprocal grooming partner preferences in an agent-based model of macaque groups

Abstract

Whether and how primates are able to maintain long-term affiliative relationships is still under debate. Emotional bookkeeping (EB), the partner-specific accumulation of emotional responses to earlier interactions, is a candidate mechanism that does not require high cognitive abilities. EB is difficult to study in real animals, due to the complexity of primate social life. Therefore, we developed an agent-based model based on macaque behavior, the EMO-model, that implements arousal and two emotional dimensions, anxiety-FEAR and satisfaction-LIKE, which regulate social behavior. To implement EB, model individuals assign dynamic LIKE attitudes towards their group members, integrating partner-specific emotional responses to earlier received grooming episodes. Two key parameters in the model were varied to explore their effects on long-term affiliative relationships: (1) the timeframe over which earlier affiliation is accumulated into the LIKE attitudes; and (2) the degree of partner selectivity. EB over short and long timeframes gave rise to low variation in LIKE attitudes, and grooming partner preferences were only maintained over one to two months. Only EB over intermediate-term timeframes resulted in enough variation in LIKE attitudes, which, in combination with high partner selectivity, enables individuals to differentiate between regular and incidental grooming partners. These specific settings resulted in a strong feedback between differentiated LIKE attitudes and the distribution of grooming, giving rise to strongly reciprocated partner preferences that could be maintained for longer periods, occasionally up to one or two years. Moreover, at these settings the individual's internal, socio-emotional memory of earlier affiliative episodes (LIKE attitudes) corresponded best to observable behavior (grooming partner preferences). In sum, our model suggests that intermediate-term LIKE dynamics and high partner selectivity seem most plausible for primates relying on emotional bookkeeping to maintain their social bonds.

Keywords: Agent-based model; EMO-model; Emotional bookkeeping; Long-term affiliative relationships; Macaques; Reciprocity; Social bonds; Socio-emotional memory.

Figure 1. Satisfaction and LIKE dynamics.

(A) Shows the decrease of the level of LIKE_ij, starting from initial value of 1.0, for different settings of LHW (solid line: LHW = 0, long-dashed line: LHW = 180 (half-life = 3 HOURs), dashed line: LHW = 720 (half-life = 1 DAY), dot-dashed line: LHW = 5,400 (half-life = 1 WEEK), dotted line: LHW = 21,600 (half-life = 1 MONTH)), given that individuals i and j do not have affiliative interactions during the next 6 weeks. Higher LHW imply a slower decrease of LIKE, a higher half-life of LIKE and, thus, slower LIKE dynamics. Note, that when LHW = 0, LIKE_ij decreases very quickly from 1.0 to 0.0 (within 50 MINUTES). (B, C) show two examples of partner-specific satisfaction and LIKE levels in one individual i due to grooming bouts received from five different partners j (B: LHW = 180, C: LHW = 720). Each partner is depicted in a different shade of blue.

Figure 2. Average LIKE attitudes.

This figure shows the distribution of LIKE attitudes among the individuals of a group for different settings of partner selectivity (LPS) and LIKE dynamics (LHW). Higher LPS defines a higher preference for affiliation partners with high LIKE attitudes. Higher LHW (Like history weight) defines slower LIKE dynamics. LIKE attitudes are directed from actors (y-axis) to receivers (x-axis), both are ordered by dominance strength, ranging from low-ranking (myDOM = 0.05) to high-ranking (myDOM = 1.00). The plot shows the LIKE attitudes of one example run, averaged over one YEAR. Dark shades represent high LIKE attitudes. Values at the diagonal are by definition not applicable.

Figure 3. Grooming rates.

(A) This figure shows the group mean of the individual grooming rates for different settings of selectivity (LPS) and LIKE dynamics (LHW). The box-plots show individual grooming rates (in minutes per hour) of 10 simulation runs, averaged over one YEAR. (B) This figure shows the distribution of the dyadic grooming rates within the group for different settings of selectivity (LPS) and LIKE dynamics (LHW). Grooming is directed from actors (y-axis) to receivers (x-axis), both are ordered by dominance strength, ranging from low-ranking (myDOM = 0.05) to high-ranking (myDOM = 1.00) individuals. The plot shows the grooming rates of one example run (the same run as in Fig. 2) in minutes per hour, averaged over one YEAR. Dark shades represent high grooming rates. Values at the diagonal are by definition not applicable.

Figure 4. Grooming rates, LIKE attitudes and reciprocity.

(A) Correlation between grooming rates and LIKE attitudes. This figure shows the row-wise correlation coefficient (Pearson’s r) between grooming rates and LIKE attitudes for different settings of selectivity (LPS) and LIKE dynamics (LHW). We used matrices of grooming rates and LIKE attitudes averaged over one YEAR. The correlation coefficient was calculated for each of 10 simulation runs and then averaged using a Fisher-z transformation. (B) Reciprocity of the total durations of grooming partner preferences. This figure shows the group-level reciprocity of the total durations of grooming partner preferences for different settings of selectivity (LPS) and LIKE dynamics (LHW). Group-level reciprocity is measured as Kendall rowwise tau. Higher tau_rw values indicate that stronger reciprocity of preferred partner preference. Taurw values were calculated based on the total partner preference durations averaged over 2 YEARS. The box-plots show the tau_rw values of 10 simulation runs.

Figure 5. Duration of grooming partnerpreference bouts.

(A) This figure shows the group mean of the average preferred partner bout durations for different settings of selectivity (LPS) and LIKE dynamics (LHW). The box-plots show the duration (in MONTHS) of 10 simulation runs, averaged over two YEARS. (B) This figure shows the distribution of the average preferred partner bout durations among the dyads of a group for different settings of selectivity (LPS) and LIKE dynamics (LHW). Partner preferences are directed from actors (y-axis) to receivers (x-axis), both are ordered by dominance strength, ranging from low-ranking (ID = 1) to high-ranking (ID = 20) individuals. The plot shows the average duration of grooming partner preferences of one example run (the same run as in Fig. 2) in MONTHS, averaged over two YEARS. Darker shades represent longer preferred partner bout durations. Values at the diagonal are by definition not applicable. (C) This figure shows the group mean of the total preferred partner durations for different settings of selectivity (LPS) and LIKE dynamics (LHW). The box-plots show the duration (in months) of 10 simulation runs, averaged over two YEARS.