EVALUATING THE SOCIAL AND TECHNOLOGICAL BENEFITS OF AN INTERGENERATIONAL PROGRAM FOR OLDER ADULTS

Intergenerational programs are on the rise. While studies have shown benefits to participation in these programs, most of the research has been focused on students and less on older adult outcomes (see Andreoletti & Howard, 2016 for a review). Currently, the University of Rhode Island (URI) is host to the Engaging Generations: Cyber-Seniors Program, which connects older adults with undergraduate technology mentors. The aim of this study was to evaluate the outcomes of intergenerational programming participation for older adults related to social isolation, loneliness, social engagement, and digital competency, as measured by the pre/post surveys given to participants. SPSS software was used to conduct descriptive analysis, paired-sample t Tests, independent sample t Tests, and one-way ANOVAs. Thematic analysis was used for the open-ended participant response. Results showed significant improvements on items of the digital competence scale, particularly in relation to social media and for those who started with lower levels of digital competence. Qualitative analysis showed that the older adults valued the technological knowledge gained, and the pleasant interactions and pedagogy. Program implications and suggestions are discussed.

iii ACKNOWLEDGMENTS I would like to thank several individuals for guiding and supporting me throughout this process. Without them, this would not have been possible.
First, I would like to express my gratitude to my major professor, Dr. Skye Leedahl, for taking me on as her first thesis mentee. She continually motivated me, pushed me forward, and helped me bring my vision to life. I cannot express my gratitude for all the time and effort she put in guiding and teaching me. I would like to thank her for all her patience and contributions, start to finish. Thank you for letting me take part in your baby (CyberSeniors).
I would like to thank my inside committee member, Dr. Phillip Clark, for his continued support. I did not come into URI thinking I would be working with older adult populations, but Phillip Clark has a way of making you excited about aging. Due to his suggestions, I was able to present a completer picture of my results. Thank you for reminding me to never forget the voices of the participants. Thank you for the laughs in times of uncertainty.
I would like to thank my outside committee member, Dr. Molly Greaney, for her unique input and suggestions. She was pivotal in understanding importance insights I would have missed. I am most grateful for all the detailed edits she provided.
Thank you for your eagerness to help and commitment to my work.
I would also like to thank my outside chair, Dr. Melanie Brasher, for agreeing to mediate my defense. She committed without knowing me, and I am eternally grateful. You were the final piece in this crazy process and your kindness took the stress right out of it.

Statement of the Problem
Under the assumption of an existing and ever increasing generational gap, intergenerational service-learning projects and courses are being implemented to connect students and older adults. While studies have shown benefits to participation (see Andreoletti & Howard, 2016 for a review), most research has focused on student outcomes, and the few assessments of older adult outcomes have found neutral to mildly positive results (Roodin, Brown, & Shedlock, 2013). Through an evaluation of the University of Rhode Island's (URI) Engaging Generations Program: CyberSeniors, a program that connects university students with older adults for student-led technological instruction, the aim of this study is to determine if participating in an intergenerational program reduces social isolation and loneliness, and increases social engagement and digital competency for the older adult participants.

Justification for and Significance of the Study
For decades, intergenerational programs have been used to foster collaboration, promote unity, and nurture cultural and community preservation between generations (Kaplan, 1997;Newman, 1997). However, despite the growth in programs, in-depth evaluations of these programs remain scarce (Kuehne, & Kaplan, 2001) and often focus on student outcomes instead of older adult outcomes (Roodin et al., 2013). A couple of studies that focused on the program impact on older adults noted reduced depression and negative self-perceptions (Hernandez & Gonzalez, 2008), and more open-mindedness towards younger generations (Young & Janke, 2013). White et al. (2002) reported a trend toward decreased depression and loneliness, but the change was not significant. In a review about intergenerational programs, older adults spoke to having the opportunity to pass down wisdom (Newman & Hatton-Yeo, 2008), share life experiences, and gain cross-generational understanding (Underwood & Dorfman, 2008). Building on the results of previous studies on intergenerational programs, this study is focused on evaluating the URI's Engaging Generations Program: Cyber-Seniors' impact on social and technological outcomes for older adults.

REVIEW OF LITERATURE
Definitions Service-learning. URI's Engaging Generations Program is implemented through service-learning (Underwood & Dorfman, 2008). Bringle and Hatcher (1996) defined service-learning: as a credit-bearing educational experience where students participate in an organized service activity that meets community needs and also provides an opportunity to reflect on the service activity in such a way as to gain further understanding of course content and an enhanced sense of civic responsibility (p. 222).
As implied by its name, service-learning is meant to enhance course material through completion of a related service, with learning for students and benefits for those receiving services being emphasized (Furco, 1996). Young adults participating in service-learning have shown increased ageism sensitivity and more positive attitudes towards older adults, particularly in regards to working with them (Augustin & Freshman, 2016 Andreoletti & Howard, 2016). This approach provides the opportunity for younger adults to practice leadership skills and for older adults to learn new skills usually associated with youth (Murphy, 2012), such as social media.
Social isolation. Social isolation is defined as the lack of integration into available social networks and supports. In a qualitative study of 30 older adults, half of the participants brought up themes of exclusion (Bell & Menec, 2015), suggesting that worries about social isolation are common among older adults. Research has found social isolation to be a risk factor for poorer physical and mental health (Miyawaki, 2015), including an increased risk of developing Alzheimer's disease (Wilson et al., 2007), higher mortality risk (Holwerda et al., 2012), and reduced cognitive functioning (Caciopp & Hawkley, 2009). Older adults are at an increased risk of social isolation due to their decreasing social networks through the loss of loved ones and friends associated with aging (Singh & Misra, 2009), and decreased mobility due to the increased chance of disability and disease (Rantakokko, Mänty, & Rantanen, 2013).

Loneliness.
There are various definitions of loneliness in the literature. In different studies, loneliness has been defined as an unpleasant and unwelcome feeling (Hauge & Kirkevold, 2010), and a painful feeling that occurs when one is not as socially or as intimately connected as desired (de Jong Gierveld & van Tilburg, 2006;Perlman & Peplau, 1981). Gerontological studies conducted in Great Britain found that 35 -46% of older adults ages 65 and older reported feelings of loneliness some of the time to most or all of the time (Cann & Jopling, 2011). In older adults, loneliness has been shown to be significantly associated with depression and suicidal ideation, particularly for minority groups and females (Wright-St Clair, Neville, Forsyth, White, & Napier, 2017).
Social engagement. Glass et al. (2006) defined social engagement as the "performance of meaningful social roles for either leisure or productive activity" (p. 606). Mendes de Leon and colleagues (2003) found social engagement to positively influence health outcomes for older adults in the areas of healthcare expenditures, disability, and mortality. Glass and colleagues (2006) found longer survival rates and reduced declines in cognitive function for older adults who were more engaged socially. Social engagement is found to have a positive effect on cognitive performance due to the increased presence of activities that exercise cognition (Brown et al., 2016). In addition, social engagement as a protective factor for cognitive functioning was particularly significant for group-based engagement, which strengthened as age increased (Haslam, Cruwys, & Haslam, 2014 As technology becomes more integrated into everyday life, digital competence is increasingly important for older adults (Czaja et al., 2006). Unfortunately, older adults are unable to learn at the rate technology is developing (Charness, Schumann, & Bortiz, 2002). In addition, computer anxiety is an obstacle to digital literacy (Laguna & Babcock, 1997). However, technology training can mitigate this anxiety (Czaja et al., 2006), improve computer skills, increase usage, and foster social connectedness and social participation (Gardner, 2010). Hampton and colleagues (2011) found increases in social ties, social support, and diversity in social networks with social media use, however, the noted that older adults are less likely to use social media.
Theoretical Framework: Engagement Theory Kearsley and Shneiderman (1998) believed that engagement theory could be used as a framework for teaching and learning through technology. According to engagement theory, for meaningful learning to occur, learning activities must be social and worthwhile to the student. Summarized by Relate-Create-Donate, learning activities must follow three components: 1) a group context, 2) project-based, and 3) an authentic focus. According to the authors, a group context or Relate, increases learning motivation, exercises social-emotional competencies, and provides opportunities to interact with diverse perspectives. Project-based learning allows for real-world problem solving and the potential for student-controlled learning. Last, an authentic focus, such as tying student learning outcomes to an outside beneficiary, can increase satisfaction and motivation. While not a key component of engagement theory, the authors believed engagement could be facilitated through technology.
For older adults, engagement theory and technological learning seems to relate to modernization theory ideas, as discussed in gerontological literature (Cowgill, 1974 As older adults are slower to learn (Charness, Schumann, & Bortiz, 2002) and less willing to use (Hampton et al., 2011) new technology, older adults may be unable to reconcile their reality with this new societal definition of progress. Marginalized, social and community engagement in older adults may be reduced, while social isolation and loneliness increased (Hooyman, & Kiyak, 2011), suggesting that technology training may alleviate social isolation and loneliness, while increasing social engagement in older adults. This paper will further expand the use of engagement theory with the focus on older adult learning, a novel population.

Program Description
Inception Purpose. The purpose of the program is to meet three objectives: "(1) promote civic engagement and service-learning for college students; (2) help prepare future health and human service professionals for careers; [and] (3) improve social connectedness and interest in technology for older adults" (Leedahl et al.,in press,p. 9 sessions, was a student-led class held once a week using iPads provided by the program. Classes were tailored based on interests of those taking the class and were held at PACE. While a student leader ran the course, other students also working on their service-learning hours would assist.

Student involvement.
This program was designed to be a learning opportunity for future health and human services professionals and a supplemental program for those with interests in geriatrics and gerontology. Students participate in the program through independent study credits, coursework, or experiential education hours. This flexibility allows for a variety of departments to be involved and for students to receive credit for their participation. Before beginning the sessions, each studentmentor takes part in an hour-long training session encompassing program logistics, tips, and problem-solving.

Current Study
The goal of this study was to advance the literature focused on the older adult outcomes from participation in an intergenerational program implemented using reverse mentoring and connecting generations through technology. The purpose of this study was to understand if and how the program benefits older adult participants.
This study examined the following evaluative research questions: • RQ1: What were the demographic and social characteristics of participants in the Engaging Generations: Cyber-Seniors program during the study period of

Design
The aims of the study were to examine who participated in the program, determine if there was a statistically significant increase or decrease between pre/posttest scores on selected measures, and to thematically analyze perceived impacts of the program. First, the study examined descriptive statistics. Second, utilizing a pre/post intervention design, this study evaluated pre/post differences on scores from loneliness, social isolation, social engagement, and digital competency measures (outcomes) after exposing older adults to intergenerational reverse mentoring (intervention). Third, this study conducted qualitative analysis using thematic analysis.

Sample
Between Fall 2016 and Summer 2017, 123 older adults participated in at least one of the models of the program, and of these, 82 participants (66%) completed the pre-survey only (non-completers) and 41 older adults (33%) completed both pre-and post-surveys (completers). All older adult participants were given the pretest survey (T1) at the first meeting between the adult and student. After the completion of three sessions, all participants were asked to take the posttest survey (T2). Importantly, 72 older adults did not participate in three sessions, and for the 51 older adults that did, 10 chose not to complete the posttest survey. The response rate for those that completed the pre-and post-surveys was 80.4 percent.

Data Collecting Tools
The research team conducted a pilot study during the Spring 2016 semester.
Then, starting in the Summer of 2017, the research team modified the surveys to include the measures that were a part of the pilot test (i.e., social engagement, social isolation), and they also added measures for loneliness and digital competence. The pre/post surveys used the same items with the exception of additional open-ended questions on the post-survey asking participants to reflect on program effectiveness.
The following scales were included on both surveys. See Appendix A for a copy of the survey items used in this study.
Social isolation was measured using the Lubben Social Network Scale (LSNS-6) (Lubben et al., 2006). The LSNS-6 includes six 6-point Likert scale questions about family and friendships, with responses of none, one, two, three-four, five to eight, and nine plus. In this study, separate scale scores for family and friendship were analyzed, as well as the individual items on the scale and an overall sum score. In analyzing this scale, a higher item score, scale score, or summed total score indicated less isolation. For this study, these measures can be considered reliable with Cronbach's alphas ranging from .867 to .912 on the pre/post sub-scales and the overall scale score.
A social engagement scale about the frequency of engagement in social activities (e.g., visiting friends) was used in the study. The measure was derived from Glass and colleagues (2006) and has been used in previous studies (Leedahl, Chapin, & Little, 2015;Leedahl et al., in press After downloading the SPSS files, the data was merged and cleaned. Prior to conducting this study, all participants had been given an ID number. IDs were matched to the participant survey(s), and then names and any identifying information were retracted from the SPSS database to de-identify the data.
Missing data was identified as not answered (-99) or not included on the survey (-88). To address scale items not answered, mean substitution was used if at least ⅓ of items had data for a respondent (Neuman, 2011). For the pretest data, mean substitution was used for one respondent who missed one question in the friendship sub-scale, two respondents who missed one question in the family sub-scale, 18 respondents who missed 1 or 2 questions in the social engagement scale, and 19 respondents missed 1 to 5 questions on the digital competence scale. For the posttest data, mean substitution was used for one respondent who missed one question on the 1 A Confirmatory Factor Analysis was conducted for the social engagement & digital competence measures. The results of this analysis showed that separating scale items into more than one-factor would reduce Cronbach's alpha, thus scale scores were analyzed overall. family sub-scale, seven respondents who missed one question on the social engagement scale, and six respondents that missed 1 or 2 questions on the digital competence scale.
Scale scores that did not meet these criteria were not included in the pre/post analysis. In the pretests, four were excluded from the friendship sub-scale, two from the family sub-scale, nine from the social engagement scale, fifteen from the digital competence scale, and one from the loneliness scale. In the posttests, one was excluded from the friendship sub-scale, one from the family sub-scale, two from the social engagement scale, three from the digital competence scale, and one from the loneliness scale.
A total sum score was created for each of the scales. In addition, the friendship and family sub-scales were combined to create a LSNS-6 (social network) sum score, and the sum scores for each sub-scale was used in the analysis. The sum score was then used to categorize the participant as high or low on the scale for all measures. The sum score was compared to the mean score of that scale. Those scoring above the mean were categorized as high and those below the mean were categorized as low. A change in score for each scale was also calculated. The pretest score was subtracted from the posttest score of each participant. This was calculated for each measure.
Descriptive statistics, frequencies, means, and standard deviations for all variables were identified. To assess pre to posttest changes in individual items and scale scores, paired-sample t tests were used. Effect size was also examined using Cohen's d. Follow-up independent t tests and ANOVAS were conducted to assess the characteristics of the participants with significant changes from pre to post scores.

Qualitative analysis. To assess perceived impacts, open-ended responses for
"what was your favorite part of the program, or the most valuable thing you learned?" were analyzed using thematic analysis. Thematic analysis reduces qualitative material to meaningful patterns or themes (Patton, 2002). In conducting qualitative analysis for this study, all responses were first gathered in a single excel document. Responses were initially read to gain an overall sense of responses. Responses were carefully read again and coded, and each response could have multiple codes. Code patterns were identified, and similar codes were collapsed into one theme. A list of five themes was identified, along with key quotes. Once a list of themes was developed, it was reviewed with Dr. Skye Leedahl, the principal investigator, and finalized by collapsing a few of the categories. For example, Advanced Use of Technology was incorporated into the Help with Use of Technology category because it represented one extreme end of the spectrum and not a separate category. An 'Other' category was established for some of the responses that did not fit into any other category.  Table 1 for demographic data for the older adults who participated in the program. Eighty-two of these participants (M = 73.57, SD = 6.74) did not complete the post-survey (non-completers). Of these 82 participants, 39 adults (48%) participated via senior center appointments, 25 adults (31%) via OLLI drop-in sessions, 14 adults (17%) via the matching program with URI classes, and 4 adults (5%) via the PACE class program.
As shown in Table 1, non-completers were primarily female (56%), White (83%), married (39%), and living with someone (45%). Most participants reported being in very good health (27%) to excellent health (26%), and a small minority were in fair (5%) to poor (1%) health. Seventy-two percent of participants were retired, 10 percent were unemployed or unable to work, and 5 percent were employed full or parttime. Over half (55%) reported an income of $30,001 or more, and 27 percent had reported incomes less than $30,000. Overall, participants were mostly well educated; 29 percent graduated college and 29 percent received a graduate degree, while only 3 percent did not complete high school.
Fifty-one older adults (41%) completed at least three sessions, and of those, 41 adults (80%) completed both the pre-and post-surveys (M age = 74, SD = 7.85). Of those who completed both pre-and post-surveys, 12 adults (29%) participated via senior center appointments, 3 adults (7%) via OLLI drop-in sessions, 17 adults (42%) via the matching program with URI classes, and 8 adults (20%) via the PACE class program. This sample did not follow the same program model distribution as the noncompleter sample. Given the format of the different program models, it was easier to get participants from certain models to take the surveys. For example, the matching program group was a higher proportion of the sample. Participants were sent an email link and asked to complete the surveys, and prior to signing up for the matching program, they were told they needed to complete at least 6 hours in the program to participate. At other sites, participants could take part in as many or as few sessions as they wanted, for example, at the OLLI drop-in sessions, sessions were designed to be quick and resolve specific questions. Multiple sessions were not necessary or required, thus getting post-survey data was more challenging.
Overall, the completer sample followed the same demographic distribution as the larger non-completer sample; female (56%), white (85%), married (42%), living with someone (49%), and in very good health (17%) to good health (17%). Eighty-two percent of participants were retired, and 59 percent reported an income of $30,001 or more. However, participants had less education; 19 percent less received a graduate degree, and 6 percent more did not complete high school. Therefore, non-completers and completers matched relatively well demographically, with those with less education perhaps participating more often than those with more education, thus explaining the educational differences between samples.
Participants also were asked about the technological devices owned on both surveys. At baseline (n=123), the majority reported owning smartphones (64%), and laptops (57%), however, a larger portion of the matching program participants owned a smartphone (82%) and laptop (74%) as compared to, for example, the class session participants at PACE (42% and 33%, respectively). In the post-survey data (n=41), tablets (58%) and smartphones (58%) were the most owned devices, and once again, Participants also were asked for what purposes the technological devices were used. In both the pre and posttests, email (76% and 80%) and searching the internet for information (60% and 80%, respectively) were the most common reported purposes.
Once again, proportions varied by program model. For example, in the pretest, only 39 percent of individual appointment participants searched for information on the internet as compared to 87 percent of the matching program participants.

Research Question 2
Paired-sample t tests were conducted to evaluate if any pre/post differences could be detected based on program participation for participants on the following measures: LSNS-6, loneliness, social engagement, and digital competence. See Table   2  Follow-up analyses were then conducted using the high/low variable categorizations. Independent samples t tests were conducted to test whether having a high or low initial pretest score would be significantly associated with the amount of change in the scale occurring pre to posttest. The tests were significant for the digital competence scale, overall and for all individual items except Using Video Calls (see Table 4). The average change for participants starting with higher digital competence (M = -5.51, SD = 8.37) was significantly less than those that started with low digital competence (M = 14.96, SD = 18.74), and in fact went in the opposite direction. Those starting with higher digital competence reported less digital competence in the posttest. This was the same trend for all significant individual items. Independent samples t tests were conducted to test whether any dichotomous demographics were associated with the amount of change in the scale occurring pre to posttest. Significant relationships were found with Using Video Calls (see Table 5

Research Question 3
Of the 51 participants who completed the post-survey, 48 (94%) replied to the open-end question: What was your favorite part of the program, or the most valuable thing you learned? Three participants did not respond. From the responses, five major themes were identified: help with use of technology, appreciation of the student teaching approach, enjoyment with the intergenerational interaction, assistance with overcoming anxiety or fear, and other (Table 8). Two additional participants were helped to set up websites. Last, two participants mentioned that there was a need to try to keep up with current technology.

"Realizing how much there is to learn about the new technology but understanding you have to keep at it to keep up. One session a week won't help if you don't utilize what has been taught or shown to you." -Matching program participant
Appreciation of the student teaching approach. Thirty-three percent of the 48 participants touched upon an appreciation for the student's teaching approach in two categories: program characteristics and student characteristics. Regarding the program characteristics, participants valued a variety of lesson styles; a conversation partner, lecture, discussion; and valued the benefit and "ease" of 1-on-1 learning.
Some participants viewed the learning as reciprocal and one participant "enjoyed the opportunity to contribute to someone's education." There was, however, a single critic of the program in response to this question. While their student was very determined to help, the participant from the matching program remarked that the matching was not done necessarily with the older adult participant's needs in mind.
Eight of the participants mentioned student characteristics that they found to be valuable as part of their experience. The participants characterized their student as "knowledgeable," "polite," "non-judgmental," and "open-minded." Instruction was given with "energy" and "total dedication." One participant was grateful for the respect their student showed for herself and others. Some of the participants were grateful for the patience their student demonstrated while working with them.
"He was easy to be with and patient while teaching me how to streamline my use of the computer. We had good communication..." -Matching program participant

Enjoyment of the intergenerational interaction.
Twenty-seven percent of the 48 participants suggested that the intergenerational interaction was the most valued part of the program. As participants wrote, learning from these students was a "pleasure" and "wonderful." Many of these participants spoke of the enjoyment of getting to know and connect with their student. Other. Three of the participants did not specify any specific portion of the program and instead mentioned that they enjoyed the program as a whole. For example, their responses were: "Everything!", "The wonder of it all!", and "I enjoyed all aspects of the program." Finally, one participant wrote about learning to be more empathetic towards other older adults who may have less abilities than him/her. Over half of participants most valued the technological knowledge gained, suggesting the program's success at technological instruction. Following the Engagement Theory principle of problem-based learning, older adults were able to have meaningful learning because it was 1) technology they wanted to learn (create)

DISCUSSION
and 2) knowledge that could be transferable to their daily lives (donate). Some of the older adults enjoyed the reciprocity of learning and understood that this program was a part of the student's education, and thusly the students could be considered the "outside" beneficiary for the older adult participants, further demonstrating the component "donate". One-on-one instruction was emphasized as a positive feature of the program by the older adult participants, which is contrary to the principle of group learning (relate). Kearsley and Shneiderman (1998) suggest that the importance of group learning lies in motivating continual participation through peer participation.
The participation of these older adults was completely voluntary, thus motivating participation was not a large concern. Attrition would remain a concern, however, group learning could have prevented learning and increased attrition. Some participants mentioned overcoming fear or anxiety in respects to technology use. The one-on-one approach may have been a key factor in helping with alleviating their fear and anxiety. Participation may have been hindered in a group setting because of the potential to feel inferior to their peers or feeling forced to learn the material at a faster pace than they are ready for. In this respect, engagement theory could be expanded to be more inclusive of all types of populations.
Items showing significant improvement on the digital competence scale were most related to social media. Improvements were found in the participation in social media, including through video platforms such as Skype, and the skills of posting to and sharing on social media. Using aspects of social media is more possible on a daily basis, in comparison to the other techniques of digital competence asked about in the measure. For example, participants were asked about searching for information about goods or seeking health information. Social media use could be a daily activity while the need for the latter searches may not come up often. The opportunity to use these newly learned skills may not have occurred in the time between pre and posttest. Yet, even if they had, it is likely they occurred at a lower frequency. Improvements in the basics of copy and paste would facilitate improvements in sharing and posting on social media, thus the adjacent improvements are complimentary.
However, the increase in social media knowledge and usage, surprisingly, did not transfer to an increase in social networks, as a whole or within the family and friendship sub-scales. One possible explanation is that social media may help with staying in touch with people, but may not necessarily expand a person's social networks or facilitate feelings of closeness, core concepts in the LSNS-6 scale. Social media may not influence these aspects, but instead changed the ways and frequency of communication (Raghavendra, Newman, Grace, & Wood, 2015), neither of which are addressed.
The program was unable to seemingly influence social isolation (as measured by changes in social networks), social engagement, and loneliness. While significance requirements were not met, the effect sizes point to some change in the loneliness (d = .29) and digital competence (d = .39) scales in the direction hypothesized. Effect sizes were even larger, ranging from .50 to .64, for significant individual items in the social engagement and digital competence measures. All other measure means went in the opposite direction hypothesized; with some participants testing lower in the posttest.
One possible explanation is a regression to the mean (Barnett, van der Pols, & Dobson, 2004). Participants may have overestimated their capabilities in the pretest and readjusted their responses once they had a better understanding of the concept.
There are two possible explanations for the lack of significant improvement in the overall measures. First, the program was not enough of an intervention to address these issues. The program itself is not centered on expanding social networks or creating pathways to these networks. This lack of focus is evidenced by the participants in the open-ended responses, none of which mentioned an expanded network or increase in outside interactions using knowledge gained. However, it was hoped that this social expansion would occur indirectly by directly teaching the basic skills necessary to navigate online networks. Yet given the individualized curriculum, learned skills were not the same across participants. Some participants may have learned more in depth how to use functions of social media that would facilitate their usage or ability to communicate to a larger group whereas others may not have been interested in social media at all. As social media and the growth of social networks were not the explicit focus of the mentoring sessions, social media usage needed to be expressed as an interest by the older adult. It cannot be assumed that all older adults would know what to ask for, and thus gain the same wealth of knowledge. This could possibly explain why more females perceived more improvements in making video calls. Females may have been more likely to ask for such assistance, as compared to the male participants.
Second, the program may be causing changes in the older adult participants, but the scales may be inappropriate measures. Cook and Campbell (1979) state that "inadequate preoperational explication of constructs" (p. 65) could hinder construct validity. Definitions of the research construct inform program activities or manipulation, and measures. The connection between the constructs, the activities, and the measures is questionable. For example, while these older adults were more knowledgeable of videos calls, the frequency of how often these videos calls were made was never addressed. More comfort with this skill does not necessarily mean more usage of the skill outside of the learning environment. The timeframe from pre to post-surveys may not have been sufficient for learned skills to be used in a measurable way. Changes in social isolation may be a longer-term outcome than the timeframe of the program, and the current short-term measure may not predict this long-term outcome well (Schanzenbach, 2012). In a review of social isolation interventions in older adults, the duration of successful interventions was between 8 weeks to 5 years, with most programs lasting less than a year (Cattan, White, Bond, & Learmouth, 2005).

Suggestions for the Program
For the future, a targeted intervention may be worth considering. Within the digital competence scale, those that started with less knowledge made more significant changes as compared to those that started with more knowledge. In fact, those with more knowledge seemed to regress. This may have been an issue with the measures; while those with more knowledge were learning, they may have been learning things beyond what was asked on the measure. This program may be better suited for those initiating the program with less knowledge. In addition, significant positive change in use of videos calls was most likely for female, low-income, and class sessions or individual appointments. It would be interesting for future studies with a larger sample to test if this trend remains and appears on other items of the digital competence measure. Lastly, further analysis into why certain changes are occurring with some program models but not others would help inform if all program models are necessary.
Despite more than half of participants directly mentioning the positive influence of the program on their ability to use technology, it seems that the connection between technology and social networking was lost. It cannot be assumed that older adults would understand their device well enough to optimize networking or even ask the correct questions to reach optimization. One suggestion for future program implementation is a more deliberate connection between technology and social networking on the part of the student mentors. For example, for the older adults who were taught how to use email, did the students show them how to find address Finally, as this was not a focus of this study, the student outcomes need to be examined. As the extensive intervention was more focused on the students, there may be more significant outcomes for this group than for the older adults.

Limitations
As this is not a true experiment, but instead a one group pretest/posttest quasiexperimental design, causal inference cannot be determined, and extraneous variables cannot be fully controlled. For example, the number of sessions attended, who participated, and what was learned was not controlled. A control group was also not included, but future research could do so. In addition, the number of sessions had different expectations per model. The matching participants were encouraged to sign up for six sessions, while the structure of the OLLI drop-ins only required one session.
Thus, while it can be stated that program participation and changes in the measured variables are associated, it cannot be claimed that program participation was the cause of this change over other variables.
In addition, older adults are not required to complete the surveys as part of their participation in the program. This affected the post-surveys in particular. Some programs, for example the matching program, were more apt at getting both pre and post-survey data. The final pre/post sample was small, and therefore, this situation limits the generalizability of the findings and the items that did trend toward significance in the hypothesized direction may have been significant had the sample been larger. Results from this analysis should be taken as preliminary findings to be further investigated once a greater sample size is available. Finally, OLLI members, in both the drop-in sessions and matching program, were fundamentally different from participants in the class sessions and individual appointments. OLLI members reported higher initial digital competence. This could explain the differences in success between models, as opposed to the actual structure of the model itself. Further analysis is recommended.

Conclusion
Qualitatively, the older adult participants are responding favorably to the program. As many did mention advancing in technological abilities and their enjoyment of the interaction, the underlying features of the program, intergenerational interaction and technological instruction, are working. However, researchers need to better match the actual goals related to social isolation and social engagement of the program to program activities and/or measures. The influence this program has on the older adults may be better measured once the more ideal quantitative measures are created and/or implemented. a Sample sizes vary across the different measures due to the listwise deletion of missing data b Scores range from 0-15, with higher scores indicating more closer friends c Scores range from 0-15, with higher scores indicating more closer family members d Scores range from 0-30, with higher scores indicating more close friends and family members e Scores range from 0-27, with higher scores meaning more social engagement f Scores range from 11-55, with higher scores indicating more competency in tasks involving technology g Scores range from 3-12, with higher scores meaning more loneliness * p < .05 ** p < .01 *** p< .001 [-3.84, -0.16]* Note: An asterisk indicates that the 95% confidence interval does not contain zero, and therefore the difference in means is significant at the .05 level using LSD