In the late 90s, Meg Ryan and Tom Hanks starred in the romantic comedy, “You’ve Got Mail”, which romanticized the idea of online dating and amplified AOL’s fame.
“The Dating Game”
Shortly after this movie, the year 2000 was the official start to dating websites with the release of eharmony, according to the “history of online dating”. Since then, the web has come alive with all types of dating websites and apps, including Match, Bumble, and Tinder. Whether you are looking for love or a fling, there is an online dating site for just about any combination of desired characteristics you might imagine. Be sure to swipe right! (Ask your kids to explain if you don’t know what that means.)
So how do online dating and autobiographical memory relate to one another? Well, let’s go back in time to the year of 1996. What did you do that May, specifically May 10th and May 18th?
“The Newlywed Game”
For me, these dates represent two significant life events – college graduation and my wedding day, respectively. Research suggests that we have more accurate recall of personal memories that are recent and/or emotionally charged.
Although these events were over two decades ago, I can still remember certain elements from each day with confidence, such as how I felt, what I was wearing, where I was, who was with me, and what we did. My husband, on the other hand, would more likely remember what happened on May 18th, but not May 10th, since that was not his college graduation date. However, if the two of us were asked about May 10th 1997 or even May 18th 1997 (or about those dates in 2019), we would both have difficulty recalling specific memories, which would make us unworthy opponents on the “Newlywed Game”!
While a first-year anniversary may seem memorable, in our case, it was over 22 years ago. Even less memorable and distinctive, May 10th, which falls around Mother’s Day, was certainly not a significant day 22 years ago as we did not have any children then. Unfortunately, my more current recall of these two days, which occurred only 3 months ago, would likely be faulty and seem like 20 years ago, despite having the consistent and distinctive cues of Mother’s Day and my institution’s commencement ceremonies.
So much for autobiographical memories in general. Now how does this relate to email?
“Family Feud”
In a recent article published in the Psychonomic Society’s journal Behavior Research Methods, researchers Abhilasha Kumar, Sudheendra Hangal, and Allyson Rosen investigated the role of email in autobiographical memories. Kumar 
and colleagues asked if recollection of significant events discussed through e-mail could be recalled accurately, and whether this ecologically more contemporary method would produce similar outcomes as laboratory-based research on autobiographical memories?
To assess both questions, the researchers developed a Web-based program that analyzed participant e-mail archives from the last year. Participant e-mails, limited only to gmail accounts, were retrieved and cleaned using a textual analysis algorithm, the MUSE algorithm. E-mails were stripped to include just the written content between the participant and another party, which could not be a generic recipient such as listservs or product companies.
Once the e-mails were cleaned and a contact address book was created, the MUSE algorithm extracted three types of ”scores”: recipient score, sentence score, and e-mail score. Recipient score was based on the frequency of e-mails exchanged with unique e-mail addresses.
The algorithm then extracted various sentences from the e-mails that were scored according to specific parameters including emoticons ( J L ), various punctuation marks (?, !), proper names (e.g., Judy, Chris, …), words based on family status (e.g., brother, mother), travel words (e.g., airplane, trip), reflective words (e.g., trust, understand), and different types of emotion words (e.g., happy, amazing, angry, depressed). These features were utilized to create a sentence score that was used to identify sentences that could be selected as individualized experimental stimuli for each participant. The sentences with higher scores had more of these elements, suggesting a more distinctive or memorable sentence.
The third feature extracted was an e-mail score that was calculated to include sentiment content and number of e-mails in a thread.
These three scores were then summed to calculate a final score that was used to select the sentences from each of the 12 months that would be used for each participant. Like the popular game show, “Family Feud” certain responses produce better scores and were ultimately used for the respective participant. Example sentences extracted from e-mails and the various scores calculated for each feature are presented in the table below.
Let’s Play Hangman
If a participant qualified for the study (i.e., he/she had enough e-mails to enough different people with enough required features), the participant was given pairs of sentences extracted from their sent e-mail. The sentences were the participant’s cue to attempt to recall the name of the recipient of the e-mail from which the two sentences were derived.
The study also provided some additional recall cues such that the number of letters in a given name was given as the number of dashes (John Smith _ _ _ _ _ _ _ _ _). Participants attempted to fill in the missing name and then rated how easy it was to recall the name. If the participants had trouble with identifying the name, the first letter of each name was given as a hint (J _ _ _ S _ _ _ _).
In addition to attempting to recall the name and rating the ease of recall, participants were also asked to rate how vivid the e-mail conversation was on a 10-pt slider scale as well as what month they thought the e-mail was written.
At the end of the experimental portion of the task, the participants filled out a post-experiment questionnaire where they were asked to explain ”wrong” responses—that is, cases where the participant’s response did not match the “answer” as determined by the algorithm. Participants were given a pre-determined list of possible reasons for their “errors” (e.g., I feel like I should have remembered this name; I have trouble remembering this name). Finally, participants answered a few other questions about their experience.
“Survey Says. . .”
Like the TV gameshow, “Family Feud”, the results of this study indicated that participants were able to correctly recall the name of the e-mail recipient almost 50% of the time. The level of recall was influenced by frequency of e-mail contact and time since the e-mail had been received. E-mail recipients that has been e-mailed more frequently were recalled more accurately (bottom panel of figure below – High: +1SD) than less frequently contacted recipients (top panel of figure below – Low: -1SD).
Similarly, more recent e-mails produced greater accuracy in name recall than older e-mails (all three panels in figure above show decreasing trends) as well as a reported perception that the name was easy to recall (blue line/dots in the figure below). In addition to decreased accuracy over longer retention intervals, participants also reported a decrease in vividness for older e-mails.
Thus far, the results have been similar to what has been observed before in conventional laboratory settings. For example, participants were more likely to date an e-mail as more recent than it actually was, rather than as more distant (known as a telescoping error), and this pattern increased linearly over time as shown in the top panel in the figure below.
Interestingly, when e-mail sentences contained more sentiments, the participants were more likely to make a telescoping error. The researchers suggested that this interaction between sentiment and temporal dating errors was related to the perceived emotional salience of the e-mail content, which makes an event more distinctive, memorable, and easier to recall ultimately.
In contrast, although participants occasionally guessed that an e-mail from a month was farther away than it actually was (i.e., time expansion error), this temporal dating error pattern was less consistent and weaker.
Alexa, what did I do on August 10, 2016?
The results of this novel, ecologically valid (and contemporary) methodology controlled demand characteristics of previous autobiographical memory laboratory-based research, such as awareness of the study’s purpose due to explicit instructions and opportunities for rehearsal.
In this study, participants responded to stimuli that were designed without their active input aside from providing their archived emails. Because the MUSE algorithm selected specific sentences from each participant’s personal email archive, the recall cues provided were personalized for each participant without biasing the selection of events. In typical laboratory-based research, participants usually provide information about relevant events in response to prompts (i.e., describe your last birthday party, what do you remember about your high school graduation). These questions can lead participants to become aware or more conscious of specific events, which could then lead to increased rehearsal of the event and distort memory for the original experience.
This novel approach may prove to be an effective tool for studying autobiographical memory within other real-world contexts, such as the memory storage and recall capabilities of Alexa, Amazon’s virtual assistant technology. Although I am one of the (seemingly) few who does not talk to Alexa on a daily basis, a quick search of my e-mail held the answer to my last question above. Apparently, August 10, 2016 was the day that I agreed to become a digital content editor for the Psychonomic Society.
Psychonomics article focused on in this post:
Kumar, A. A., Hangal, S., & Rosen, A. C. (2019). Autobiographical recall of personally familiar names and temporal information in e-mails: An automatic analytic approach using e-mail communications. Behavior Research Methods. DOI: 10.3758/s13428-018-1182-9.
