Social media has become a platform for new mothers to openly share their experiences of the joys and challenges of parenthood. Researchers at Dalhousie and Penn State have used artificial intelligence to unravel the sentiments in nursing mothers’ tweets to better understand the factors influencing breastfeeding behaviors. They hope the findings can inform policies and interventions to support and improve resources for nursing mothers, such as breastfeeding support, workplace accommodations and technological aids such as apps.
“This research is very important for many reasons: attracting and retaining women in various roles and closing the gender gaps that exist in many professions cannot be actualized without addressing the unique issues that women face; providing enabling environments, interventions, and policies that support women through various life stages such as during breastfeeding,” says Dr. Rita Orji, Canada Research Chair, and associate professor in the Faculty of Computer Science. “It is also important for improving maternal and infant health. Giving birth and breastfeeding is a life-changing experience that can cause issues including depression and stress. Adequate support from different stakeholders – family, workplace, government, and society at large is needed as they navigate this process. It is surprising that even in the 21st century, women still feel largely unsupported and uncovering these issues is the first step towards finding the appropriate solutions.”
Breastfeeding themes
In the study, researchers collected more than 19,000 breastfeeding-related tweets. Using existing lexicon-based tools and new machine learning classifiers that they developed, they classified the data to predict the sentiment polarity — whether the text was positive or negative — of the behavior described in each tweet.
“We are getting the raw sentiment of nursing mothers without putting them in a controlled experiment environment where their views could become biased,” says Dr. Richard Lomotey, assistant professor of information sciences and technology at Penn State Beaver. “We are getting the real frustrations and joys experienced by these lactating mothers, which can help us to really explore the questions on the ground and the interventions that can be proposed to assist in this regard.”
The negative issues tweeted about by nursing mothers run the gamut, from latching problems and low milk supply, to postpartum depression and lack of support, to criticism over public breastfeeding. Positive tweets highlighted perceived benefits of breastfeeding, such as mother-child bonding, nutritional value, and access to breastfeeding resources.
In all, 29 negative themes and 21 positive themes were identified. While most of the factors affecting breastfeeding behaviors negatively have been reported in existing literature — such as latching problems and short maternity leave — the researchers’ analysis uncovered some new factors that have not previously been studied, including the deliberate decision to not breastfeed and fear of biting.
“Breastfeeding is a very popular issue, but you don’t see public health officials and policy makers discussing it publicly or public engagement on this topic,” said Lomotey. “Yet, it’s a very important issue from a health perspective. So that’s one of the driving factors in this project, to explore these issues that are being swept under the carpet and certainly considered a public taboo in some regions.”
Highlighting issues and possible interventions
The researchers then placed the themes they identified into four negative categories of health-related issues, social factors, psychological factors, and situational factors: and four positive categories of perceived benefits, maternal self-efficacy, social support, and education and training support.
“We highlighted a lot of issues that impact breastfeeding women, and we’re not going to be able to address them all,” said Rita Orji, associate professor of computer science at Dalhousie University. “But we want to put this out there so that policymakers and leaders at workplaces and schools can begin to engage with these topics and begin to highlight the issues that exist that impact women, as well as possible solutions.”
In their paper, the researchers also propose various interventions that address the negative factors to propose positive breastfeeding behaviors. For example, mothers facing social factors such as lack of lactation rooms at work could benefit from technological interventions for delivering public awareness and sensitization programs designed to increase employer’s awareness on the need to create a breastfeeding-friendly workplace.
“Many mothers have conflicting interests of a baby and a career, and this is something that fundamentally needs to be addressed,” said Orji. “We want to create a strong workforce that accommodates women and empowers them to succeed — especially while they’re breastfeeding, which is a right and a natural phenomenon in a woman’s life.”
Persuasive computing
This is one the first projects to come out of the recently launched Persuasive Computing Lab, led by Dr. Rita Orji in the Faculty of Computer Science. The lab is interested in investigating user-centered approaches to designing interactive systems to motivate people towards actions and causes that are beneficial for them and their communities, as well as how interactive systems can be designed for under-served populations. Dr. Orji’s team applies research to tackle real-life problems in various domains to improve a wide range of health and wellness behaviours.
“Our research provides us with the opportunity to design technologies that empower people and contribute to solving wide ranging problems in society,” says Dr. Orji. “While this project is based on breastfeeding, these methods can be applied to any issue that is being widely discussed on social media. An example of this can be seen in another project I am working on with colleagues at Dalhousie around responses to social isolation during COVID-19.”
Drs. Orji and Lomotey worked with Oladapo Oyebode, a research and teaching assistant in the Persuasive Computing Lab. Their work appeared in the April 2021 issue of IEEE Access, published by the Institute of Electrical and Electronics Engineers.
The collaboration is currently been funded by the Penn State Beaver Office of Academic Affairs, the Canada Research Chairs Program and the Natural Sciences and Engineering Research Council of Canada (NSERC) through the Discovery Grant.