Source-linked AI summary
Daily rhythms in mobile telephone communication
Talayeh Aledavood, Eduardo López, Sam G. B. Roberts, Felix Reed-Tsochas, Esteban Moro, Robin I. M. Dunbar, Jari Saramäki
TL;DR
Questionnaire studies capture broad morningness and eveningness, while electronic records have often been analyzed in aggregate, leaving the social dimensions of individual daily rhythms less understood. Using longitudinal mobile-call and questionnaire data, the paper finds distinct individual calling patterns that persist despite network turnover, alongside time-of-day differences in whom people call.
Problem
The social dimensions of circadian rhythms and their variation across individuals are less understood than broad morningness and eveningness or aggregate electronic-record patterns.
Source: §I · p. 1Method
The study combines questionnaire data with longitudinal mobile-phone records to examine individual call rhythms and their recipients.
Source: §I · p. 1Results
Individual daily call patterns differed and persisted despite network turnover, while evening and night calls often focused on specific, emotionally close alters.
Source: §I · p. 2, A. Persistence of individual daily… · p. 2, §III · p. 6Takeaways & Limitations
Individual differences in circadian rhythms include a social component: people direct calls to particular contacts at particular times of day.
Source: §I · p. 2The intensive longitudinal data collection resulted in a relatively small sample size.
Source: §III · p. 6
Abstract
from arXiv · showhide
Circadian rhythms are known to be important drivers of human activity and the recent availability of electronic records of human behaviour has provided fine-grained data of temporal patterns of activity on a large scale. Further, questionnaire studies have identified important individual differences in circadian rhythms, with people broadly categorised into morning-like or evening-like individuals. However, little is known about the social aspects of these circadian rhythms, or how they vary across individuals. In this study we use a unique 18-month dataset that combines mobile phone calls and questionnaire data to examine individual differences in the daily rhythms of mobile phone activity. We demonstrate clear individual differences in daily patterns of phone calls, and show that these individual differences are persistent despite a high degree of turnover in the individuals' social networks. Further, women's calls were longer than men's calls, especially during the evening and at night, and these calls were typically focused on a small number of emotionally intense relationships. These results demonstrate that individual differences in circadian rhythms are not just related to broad patterns of morningness and eveningness, but have a strong social component, in directing phone calls to specific individuals at specific times of day.
I. INTRODUCTION
Earlier work studied broad morningness and eveningness through questionnaires or analyzed daily rhythms in electronic records aggregated across people. This study combines questionnaires with mobile-call records to examine individual call patterns, their persistence, and the social relationships called at different times.
I. INTRODUCTION
- Earlier studies used questionnaires for broad morningness and eveningness or examined circadian rhythms in electronic records aggregated across many individuals.The paper takes an individual-centric approach to address this gap.Source: §I · p. 1
- The dataset tracks 74,124 calls from 24 individuals over 18 months and includes three surveys about called contacts, including their gender and kinship.Participants experienced substantial social-network turnover as they finished high school and moved to work or university.Source: §I · pp. 1–2
- Each individual had a distinct daily call-frequency pattern that persisted despite substantial turnover in their social network.The authors interpret this persistence as suggesting the patterns are characteristics of the callers rather than dependent on specific alters.Source: §I · p. 2
- Evening and night calls were typically directed to specific alters, and women’s calls lengthened toward night and were typically made to friends rather than family.The study examines daily rhythms not only in call counts but also in recipient diversity and who is called.Source: §I · p. 2
A. Persistence of individual daily call patterns
Individuals’ call patterns were more similar to their own patterns across intervals than to other people’s patterns within an interval. This difference persisted despite substantial turnover in their social networks.
A. Persistence of individual daily call patterns
- The analysis compared each ego’s average hourly call fractions across three consecutive six-month intervals, using separate intervals to assess persistence.JSD was used to compare within-ego patterns across intervals and patterns of different egos within one interval.Source: A. Persistence of individual daily… · p. 2
- Across all 24 egos, distinct daily call patterns remained similar across three six-month intervals despite major social-network turnover.The authors suggest the patterns reflect intrinsic driving forces because they do not strongly depend on network composition.Source: A. Persistence of individual daily… · p. 2
- 87% ± 12% of reference distances exceeded self-distances; mean JSD was ⟨dself⟩=0.05 ± 0.22 vs ⟨dref⟩=0.083 ± 0.28 (t=6.98, p≪10^-6).The ℓ2 comparison also found ⟨dself⟩=0.11±0.02 and ⟨dref⟩=0.14 ± 0.03 (t=6.11, p < 10^-5).Source: A. Persistence of individual daily… · p. 2
B. Alter-specificity in call patterns
Calls to the two top-ranked alters account for much of the lower alter diversity at later hours, especially at night, though patterns vary across individuals. Women’s calls were longer than men’s, particularly in the evening and at night, with differences shaped mainly by calls to friends.
B. Alter-specificity in call patterns
- Relative entropy compares observed call entropy with a weekly-shuffled reference; values below 1 indicate that certain alters are emphasized within a six-hour interval.Values near 1 indicate no alter-specificity.Source: B. Alter-specificity in call patterns · p. 2
- Relative entropy was highest in the afternoon at about 1 and lowest at night, when calls were few and typically directed to specific alters.Entropy patterns differed among egos and appeared fairly persistent, but were more variable and noisy in low-call morning and nighttime periods.Source: B. Alter-specificity in call patterns · pp. 2–3, Figure 3 · p. 4
- The fraction of calls to an ego’s two top-ranked alters generally rises toward late hours and is often highest at night, accounting for the nighttime entropy decrease.The top-alter fractions also vary across individuals and intervals; interval I1 differs in some cases during the transition from high school to summer holidays.Source: B. Alter-specificity in call patterns · p. 3, Figure 4 · p. 4
- 15 of 24 egos had significant entropy–top-alter correlations; 13 were negative and two positive, averaging r ≈−0.72.For more than half of egos, low entropy was associated with a high fraction of calls to top alters; results were inconclusive for almost all others.Source: B. Alter-specificity in call patterns · p. 3
- Women’s calls were longer overall, especially in the evening and at night and mostly to friends; kin-call durations were similar, while female-to-male calls were longest at night.The time-of-day difference was significant for 9 of 12 women but only 2 of 12 men at p < 10−6.Source: B. Alter-specificity in call patterns · pp. 4–5
III. DISCUSSION
Individual calling rhythms persist despite substantial social-network turnover, while evening and nighttime calls show gendered differences and often focus on emotionally close relationships. These patterns suggest that daily phone activity reflects both individual tendencies and the social relationships callers choose to maintain.
III. DISCUSSION
- Calling patterns differ across individuals and remain consistent across three periods despite substantial network turnover.The authors describe these patterns as extending beyond questionnaire measures of morningness and eveningness.Source: §III · p. 5
- Women’s calls are generally longer than men’s, especially in the evening and at night, with female egos’ evening calls to males and friends especially long.These calls often involved specific, top-ranked alters, who may have been boyfriends.Source: §III · p. 5
- Calls were concentrated from midday to late evening, with relatively few before midday; the authors suggest the evening pattern may extend to conversations with close friends.Why evening calls show this pattern remains unclear and merits further study.Source: §III · pp. 5–6
- Many, though not all, egos prefer to call certain alters at night, typically one or two people with special status, mainly males and friends.The authors report this pattern as especially characteristic of female egos and less so of male egos.Source: §III · p. 6
- Both sexes called kin less often in the evening, which the authors interpret as consistent with kin relationships needing less active maintenance than friendships.They note that friendship quality can deteriorate within months without sufficiently frequent contact.Source: §III · p. 6
- The study’s small scale limits its scope, while complete alter identities and relationship information provide individual detail uncommon in large-scale network studies.The authors note that frequently called alters are those with whom participants have emotionally close relationships and frequent social contact.Source: §III · p. 6
V. AUTHOR CONTRIBUTIONS
The author contributions covered study design, data collection and preparation, research conception, analysis design, analysis, and manuscript drafting.
V. AUTHOR CONTRIBUTIONS
- SGBR and RD designed the experiment and collected the data.Source: §V · p. 6
- EL parsed and pre-processed the data, while TA, JS, SGBR, RD, EL, EM, and FTS conceived the research and designed the data analysis.Source: §V · p. 6
- TA and JS performed the data analysis, and TA, JS, EM, SGBR, and RD drafted the manuscript.Source: §V · p. 6
1. Our data and its use
The study combines outgoing-call records with participant questionnaires about their social contacts to examine call patterns over time.
1. Our data and its use
- The study uses 18 months of outgoing call and text records from 24 participants, but analyzes call records to mobile phones and landlines.Source: §1 · p. 7
- Participants completed questionnaires at months 0, 9, and 18, identifying contacts as kin or non-kin and reporting relationship characteristics.Reported characteristics included gender, emotional closeness, and face-to-face contact frequency.Source: §1 · p. 7
- Itemized, time-stamped monthly bills were obtained with subscribers’ agreement, and questionnaires collected the phone numbers associated with each contact.Participants received free mobile phone accounts for taking part; anonymized, aggregated data are available online through the SI of.Source: §1 · p. 7
2. Calculating daily patterns
Daily calling patterns are calculated by grouping calls into time bins and measuring each bin’s share of calls during an interval.
2. Calculating daily patterns
- Calls are assigned to six-hour time bins by timestamp, and the count in each hour is divided by the total calls in the interval to obtain a fraction.Source: §2 · p. 7
- The calculation uses data from all days within the interval of interest.Source: §2 · p. 7
- Only complete weeks within each time interval are included.Source: §2 · p. 7
3. Measuring similarity of patterns
The section defines measures for comparing distributions and call patterns. Jensen-Shannon divergence compares probability distributions, while the ℓ2-norm is used to verify results calculated with JSD.
3. Measuring similarity of patterns
- Jensen-Shannon divergence measures the difference between two probability distributions.It is a form of Kullback-Leibler divergence that also works when distributions contain zero-valued elements.Source: §3 · p. 7
- For call distributions, π_i(a) is the fraction of calls to each alter, and H denotes Shannon entropy.The passage gives H(P) = −P p(a) log p(a).Source: §3 · p. 7
- The ℓ2-norm is used as a similarity measure between two distributions to verify results calculated using JSD.Source: §3 · p. 7
4. Entropy patterns and relative entropies
Call entropy is computed from the distribution of calls across alters within an hour or range of hours. Relative entropy compares this value with a shuffled-data reference that preserves weekly call frequencies and alter counts while removing time-specific alter preference.
4. Entropy patterns and relative entropies
- Call entropy for an hour or range of hours is computed from the fractions of calls directed to each alter during that period.The method first counts each alter's share of calls in the specified period, then computes entropy from those fractions.Source: §4 · p. 7
- Relative entropy is Horig divided by the average reference entropy from 1,000 shuffled realizations.Calls are shuffled within each week; the reference preserves call frequency patterns and calls per alter but removes preference for a particular alter at a particular time.Source: §4 · p. 7
- Weekly shuffling is used to minimize effects of long-term dynamics, including declining calls to alters and alters first appearing during the six-month interval.Source: §4 · p. 7