BACKGROUND: Being socially connected is related to well-being, and one way of avoiding social isolation is to deepen existing relationships. Even though existing relationships can be reinforced by regular and meaningful communication, state-of-the-art communication technologies alone do not increase the quality of social connections. Thus, there is a need for the involvement of a trained human facilitator in a network of older adults, preferably for a short period, to promote the deepening of their relationships. OBJECTIVE: This study aimed to evaluate the hypothesis that a human-facilitated, media-sharing social networking system can improve social connection in a small group of older people, who are more vulnerable to social isolation than most, and deepen their relationships over a period of a few weeks. METHODS: We conducted the design and evaluation of Media Parcels, a novel human-facilitated social networking system. Media Parcels is based on the metaphor of a facilitator collecting and delivering parcels in the physical mail. Extending the metaphor, the system supports a facilitator in designing time-based dialogue requesting parcels from participants that bring out their memories and feelings, in collecting the parcels, wrapping them in annotations that communicate the corresponding requests, and delivering the wrapped parcel to a target person. Qualitative evaluation was carried out in two trials with a group of three people each, one with family members (children and father; aged 55, 56, and 82 years old) and the other with a group of friends (aged 72, 72, and 74 years old), over two weeks. In each trial, data were collected in three interviews (pre-, mid-, and posttrial) and via system logging. RESULTS: Collected data indicate positive social effects for deepening and developing relationships. The parcel metaphor was easily understood and the computational system was readily adopted. Preferences with regard to media production or consumption varied among participants. In the family group, children preferred receiving media parcels (because of their sentimental value) to producing them, whereas the father enjoyed both. In the friendship group, preferences varied: one friend enjoyed both producing and receiving, while the other two preferred one over the other. In general, participants reported a preference for the production of items of a certain type depending on the associated content. Apart from having a strong engagement with the system, participants reported feeling closer to each other than usual. CONCLUSIONS: For both groups, Media Parcels was effective in promoting media sharing and social connections, resulting in the deepening of existing relationships. Its design informs researchers who are attempting to promote social connection in older adults.