Professional Documents
Culture Documents
_11/01/2011__
Amount Requested: $__75,000__ (Total amount requested) Select one choice below that applies New Application Revised App: Current ID #: ______________________ Suicide Supplement: Granting Inst:________________ Title:________________________________________
Please read the Foundation's Policies for Standard Research Grants before completing this application. Project Title (limit to 100 characters): Identifying Suicidal Indicators from Social Media Profiles and Public Posts Key Words (3 to 5 key words on target population, special methods, etc.): Profiles, Intelligent Systems
Entezari, Steven Principal Investigator (last, first) Indiana University Purdue University Indianapolis Institution Human_Computer_Interaction___________________________________________________________________________________ Department 420 University Boulevard Address--line 1 317-274-4581 Phone number PhD Informatics Degree(s)
Signature
Address--line 2 Indianapolis, IN 46202 City, State, Zip code _________________________________________________ __________________________________________________ Department head
A-1
Revised: 08/01/2011
I hereby certify that the information in this application is true and correct to the best of my knowledge. If a Standard Research Grant award is made, I understand that it is subject to the terms, conditions, and procedures stated in the AFSP Policies for Standard Research Grants. I have reviewed this document and agree to abide by all stated policies. I understand that failure to comply with these policies may result in termination of the grant. If a grant award is made, I agree to accept responsibility for the scientific and ethical conduct of the research. I understand that if the grant is not started within 6 months of award date, it will be cancelled.
_____________________________ Date
B-1
Revised: 08/01/2011
APPLICATION CHECKLIST
Please use the forms in this package for your submission. Check boxes are provided next to each application section for your convenience. All items in this application are required. If you are completing this form electronically, you may [ctrl]+click on the section headings in this list to be taken to the respective sections.
SECTION
Cover Page Assurance Form Application Checklist Abstract of Proposed Research (1 page) Budget (1 page) Budget Justification (2 pages) Biographical Information (2 pages per person) Project Description (5 pages) Project Timeline (1 page) Certification for Protection of Human and Animal Subjects (2 pages) References (no limit)
A B C D E F G H I J K L
C-1
Revised: 08/01/2011
This research will address social networking posts and profiles that illustrate the ideation of suicidal tendencies and profiles. Based on research within classical psychology, tactics to identify suicidal ideation and elicit appropriate suicide prevention techniques within the online domain are the logical next step in reaching out to those who cry for help. Our research will aim to validate the ability of intelligent systems to be designed to identify suicidal ideations that have been publically posted on a social media site, develop the criteria necessary to assess the disposition of an individual based on the individuals cry for help and their public profile page, and identify whether avenues of preventative interventions are available/feasible. We will collect data from the social networking sites Facebook and Twitter. The data collected will be based upon a review of current works on suicide identification and ideation. Metrics will be derived and validated to identify profiles that have suicidal indicators based upon the initial public profile posts. The bottom-up approach of first identifying individual posts, then assessment of the profile/persona allows situational factors to be considered by the coders and the system, when implemented. Intervention tactics will be explored, but not instantiated. The next steps will be to identify the implementation, acceptance, and efficiency of these intervention techniques. Future works will aim to identify what types of interventions can be instantiated in this new paradigm.
List all grant personnel in the table below. Include all current and/or recent institution(s) and project role(s). Key Personnel Institution Role on Project
Steven Entezari
Principle Investigator
D-1
Revised: 08/01/2011
BUDGET
Budget must not exceed $37,500 per year, or $75,000 for a two-year period. Please consult the Policies for Research Grants for information on allowable and non-allowable costs.
PERSONNEL* Name Steven Entezari John Smith Jane Jackson Role on Project Principal Investigator Graduate Student Graduate Student % Time on Project ** 80% 10% 10% SUBTOTALS CONSULTANT COSTS First Year Salary Request $37,500 $0 $0 $37,500
Dollar Amount Requested First Year Fringe Benefits $0 $0 $0 $0 First Year Sal. + Fringe $37,500 $0 $0 $37,500 Second Year Sal. + Fringe $30,000 $0 $0 $30,000
SUBTOTALS
OTHER THAN PERSONNEL First Year Totals Publications Travel Disseminate Information Conferences/ Seminars N/A N/A $0 $0 $0 $0
SUBTOTALS
$0 $0 $0
GRAND TOTAL
$37,500
$37,500
* List all key personnel and their percent time, regardless of whether salary support is requested. All personnel
who will be playing a substantial role on the project must be specifically named and biographical information included in Section G. ** The Principal Investigator must devote a minimum of 10% time to the proposed project.
F-1
Revised: 08/01/2011
PERSONNEL. Briefly describe duties and responsibilities of all key personnel. The PI, Steven Entezari, will be responsible for the project in its entirety. He will ensure that the timelines are kept to and that the board receives an update regarding the status of the project on a quarterly basis. The budget will allot for 20% of his annual salary and 33% of his summer salary (split amongst two years). Jane Smith, Graduate Student, and John Smith, Graduate Student (no relation), will be under the supervision of the PI for the duration of the project. Their duties will include content analysis basic research upon standard practices within the literature. CONSULTANT COSTS. N/A
OTHER THAN PERSONNEL. Publications in relevant journals and conference proceedings will be applied for by the principle investigator. The fees and travel costs to publish and participate in these journals and proceedings will be the subject of the budget allotted to Other Than Personnel. This will cover the principle investigators travels to and from at least two conferences, accommodations, and publication costs.
F-2
Revised: 08/01/2011
BIOGRAPHICAL INFORMATION
Provide the following information for all personnel who will play a substantial role on the project. Do not exceed 2 pages per person and do not append full CV. Name Steven Entezari Degree PhD Sex: Male
Female
1. EDUCATION (Begin with baccalaureate degree and include postdoctoral training, if applicable.) Institution and Location Purdue University, Indianapolis Indiana University, Indianapolis Degree B.S. Ph.D Year Conferred 2010 2015 Scientific Field CIT HCI
2. EXPERIENCE. List in chronological order professional positions held, concluding with present position. Research Assistant Group Psychology and Performance Lab @ Indiana University Indianapolis Adjunct Faculty Human Computer Interaction @ Indiana University Indianapolis Assistant Professor Human Computer Interaction @ Indiana University Indianapolis 3. SELECTED PEER-REVIEWED PUBLICATIONS. List (in chronological order) representative publications relevant to this application. Entezari, S. (2010). Identification of Mental Health Disorders within Social Media. In: Proceedings of ACM CHI 2010 Conference on Human Factors in Computing Systems 2010. pp. 67-89.
Entezari, S. (2011). Bystander Interventions in Suicide Management. In: Proceedings of ACM CHI 2011 Conference on Human Factors in Computing Systems 2011. pp. 67-89.
G-1
Revised: 08/01/2011
4. RESEARCH SUPPORT. List funded research projects relevant to this application (Federal and non-federal support).
Past Support: Dates N/A Source Amount Title Your Role on Project
Current Support: Dates N/A Source Amount Title Your Role on Project
Pending Support: Please list all pending applications and briefly describe below any overlap in the research proposed in any of these applications with that proposed in the present application to AFSP. Dates N/A Source Amount Title Your Role on Project
G-2
Revised: 08/01/2011
PROJECT DESCRIPTION
SECTION I. SPECIFIC AIMS/HYPOTHESES There are three main objectives for this research. The first objective is to validate the ability of intelligent systems to be designed to identify suicidal ideations that have been publically posted on a social media site. Data-mining and text-based searches of sites like Facebook and Twitter are beginning attain a set of standards and accepted practices within the marketing/research domains (Russell, 2011). Intelligent systems will be designed and tested to establish whether online cries-for-help can be isolated from the vast number of public posts that are sent out daily from these types of social networking sites. Our second objective will be to develop the criteria necessary to assess the disposition of an individual based on the individuals cry for help (as identified in objective one) and the users other publicly shared information. This other publically shared information includes the users self-generated profile information, number of connections/friends/followers and content/quantity of surrounding posts. Behavioral changes, such as timing of posts, quality/quantity of recent/previous posts messages, and the disposition of the users mood within these texts will be analyzed as well. Research has demonstrated the ability to extract a public mood from socially-posted material, like twitter posts (Bollen, Mao, & Zeng, 2010). Our third objective aims to identify whether avenues of preventative interventions are available. These avenues will be congruent with non-online intervention tactics. Establishing these avenues of intervention on the social media sites will shed light on a domain of suicide-intervention-communication that has not yet been explored. As practitioners are provided the opportunity to engage with members of the community who display these suicidal ideations, they can establish a new paradigm of crisis-response to accommodate for the virtual world. SECTION II. BACKGROUND AND SIGNIFICANCE Social networking sites, like Facebook and Twitter, allow their users to post public, text-based messages about their lives, thoughts, and actions to a group of friends/followers. Throughout this media, suicidal ideations have been posted and covered by literature in mainstream-news, but not in the professional literature (Ruder, Hatch, Ampanozi, Thali, & Fischer, 2011). Our goal is to identify suicidal ideations on social networking sites, assess the suicidal dispositions of the individuals making those statements, and explore possible interventions strategies unique to the social-networking paradigm. Users of these social networking sites expose their lives to the public by sharing micro-sized pieces of information with their online community. Within these micro-blogs, for some, like cries for help from the poster to their community. Unfortunately, however, due to the nature of online communities (i.e. psychological effects like disinhibition, deindividuation, etc) these cries for help will usually go unanswered (Marimoto & Friedland, 2010). Sites like Facebook and Twitter also give the user a sense of talking to someone. For instance, a user who posts a suicidal note on Facebook may believe many others have seen it because it was made available to the public/friends. This spotlight effect could have catastrophic effects if the user believes that their friends saw it, but did nothing. The content generated by the individual, however, can be mined and analyzed using a system capable of interpreting and categorizing the messages within the posts. Moreno et al. (2011) evaluated two-hundred Facebook profiles for depressive symptoms and criteria that, according to the DSM, categorized these individuals for MDE (Major Depressive Episode). They found that they were able to identify depressive symptoms in twenty-five percent of the profiles while also categorizing two and a half percent of the profiles into the MDE category. This result was based solely on the public posts of the participants. Our research aims to expand beyond this and venture into other self-generated content from the poster. Intelligently identifying and analyzing posted content is necessary to assess the individuals current state of mind. However, the next steps are just as crucial. Social networking sites like Facebook and Twitter have profile and other ancillary information available about the poster. This information, utilized and analyzed correctly, could give an insight into the persona of the individual as well as develop a coherent person to be analyzed. There exist methods to predict an individuals personality type simply using an interpretation of the content the user has on their social media sites (Golbeck, Robles, & Turner, 2011).
I-1 Revised: 08/01/2011
There are nearly one million suicide attempts made every year with over thirty-two percent of all suicide completions being from the age range of thirteen to thirty-four (American Foundation for Suicide Prevention). Nearly seventy-five percent of Facebook users are between the ages of thirteen and thirty-four (CVP Marketing Group, 2011). As more and more people begin using social media sites like Facebook to micro-blog information about their lives, there become more opportunities for these individuals to produce a cry-for-help for a larger audience. Many scales exist today that measure an individuals risk of suicide. There are some metrics that analyze psychometric properties of an individuals suicidal thoughts (Beck, Steer, & Ranieri, 1988) while there are others that analyze an individuals risk factors based on an assessment of the individuals current disposition (Canales, 2011). Similar to being able to assess an individuals personality type based on their social media content (Golbeck, Robles, & Turner, 2011), the establishment of a methodology to assess suicidal risk in individuals via their social media content will allow for cases that have slipped through the cracks to be identified and potentially addressed. SECTION III. RESEARCH METHODS The overall goal of this research is to address the growing problem of unaddressed suicidal ideations of at-risk individuals who present cries for help on social networking sites like Facebook and Twitter. A robust system will be developed to identify suicidal ideations and assess those individuals suicidal dispositions. Based on what is found, possible intervention tactics will also be addressed. Although this research does not aim to initiate any interventions, it does attempt to identify possible intervention tactics that could be researched in future works. The main objective of this research is to identify and assess individuals who post a suicidal cry-for-help on a social networking site. Identifying Suicidal Ideation We define a cry for help as the publications of an individuals ideations of suicide on a social networking site. According to the National Institute of Health (2011) Most suicide attempts do not result in death. Many of these attempts are done in a way that makes rescue possible. These attempts are often a cry for help. These cries for help can be identified given keywords for searches within these social networking sites APIs (Application Programming Interface). The establishment of the keywords to use will be done by reviewing the literature on previous methods of publicizing suicidal ideations including but not limited to suicide notes, first-person accounts, and interviews with individuals who have attempted suicide. Content analyses of suicide notes have also previously been studied for motivation (Osgood & Walker, 1959), verbal usage (Gottschalk & Glester, 1960), and emotion (Tuckman, Kleiner, & Lavell, 1959). We will aim to conduct our initial validity analyses from two-thousand Facebook and Twitter posts that include some ideation of suicidal tendencies. Due to the nature of Facebook and Twitter, being social networking sites, a majority of the individuals who post their thoughts, do so for the world to see. These public posts will be the subjects in our analysis. A tool will be developed to find the public posts that have suicidal ideations, as found in the first part of the research, from the content analyses. Three coders will then analyze a subset of the posts, independently judge them based on the levels of suicidal ideation present, and inter-rater reliability tests will then be performed. Any changes that need to be made to the coding scheme or development of the utility will be made, and a second run will happen with a larger subset of data. At this point, the utility itself will have the ability to judge the posts based on levels of suicidal ideation, thus, making the utility itself another rater. Another inter-rater reliability assessment will be done to ensure the most recent changes were understood by both the system and the raters.
I-2
Revised: 08/01/2011
Assessing the Disposition Once the target posts have been identified, the next step in our model is the assessment of the user profiles affiliated with the target posts. Demographic information, size of community, interaction with others, and other relevant data points will be analyzed to determine the users suicidal disposition. Similar to the content analysis of a suicidal post, this analysis will look at the content of a users profile and systematically categorize the infidel onto a suicide-risk scale. Factors outside of pure content-analysis will be examined. Certain socioeconomic, demographic, psychiatric and familial factors have been shown to be pre-determinants of suicidal risk within some individuals (Quin, Agerbo, & Mortensen, 2003). A scale to address such suicidal risk is yet to be developed for the specific domain of social media content analysis. In developing this scale, literature in related fields, such as social work and sociology, will be addressed and tested to determine their applicability to the social-networking domain. The profiles to be analyzed here are the profiles of the suicide ideation posts from the first objective. The content of these posts will not be included during the assessment of this phase; only the profile-related information of the user. Three coders will again analyze the content of these profile pages, based upon the metrics developed for the scale mentioned before. Inter-rater reliability tests will be conducted to ensure that the ideas being conveyed and the ratings of the profiles are among a standard line of thinking. The principle investigator and two graduate students will conduct the analysis. Based on the analysis, there will be a validation of the originally developed scale as well as a set of rules for the development of the new utility. Once the analysis has been completed and an acceptable inter-rater reliability has been reached, the principle investigator will begin development on the utility that will automatically rate the profiles based upon the rules from the profile analyses. Once completed, the utility itself will be utilized as another rater, and one last inter-rater reliability will be ran to validate the level of precision at which the utility will be able to assess the disposition of the users profile page. At this point, there will be three human raters and one utility acting as a rater.
Code Profile s
Review Literatur e
Develop Utility
Figure 2 - Process of Assessing Profiles Intervention Strategies Once we have identified the target posts involving suicidal ideation and assessed the profiles for disposition of suicidal risk, we begin to explore the next phase; intervention. As mentioned previously, intervention strategies will not be a major focus during this research, but more of an establishment for future works. Based on our findings with the target posts and profile dispositions, we will have successfully been able to identify current suicidal ideations (present moment), a series of risk factors that lead to this point from their profile dispositions (past moments), and potentially identify what can be done (future moments). Suicide intervention for general interventions (Tierney, 1994) as well as specific interventions, such as school settings (King & Smith, 2000), have shown increased confidence and resulted in self-reported positive results of interventions. The applicability of these methods is something to be tested within future research. However, we feel it necessary to identify some possible avenues of future interventions based upon what has worked with suicide interventions in the offline world, and what is available in the online-world. In other words, we hope to identify what resources may transfer over into the online-realm from the current suicide literature on intervention practices.
Objecti ve 2
Objecti ve 3
I-4
Revised: 08/01/2011
PROJECT TIMELINE
Month/Year August/2012 September/2012 October/2012 November/2012 December/2012 January/2013 February/2013 March/2013 April/2013 May/2013 June/2013 July/2013 August/2013 September/2013 October/2013 November/2013 December/2013 January/2014 February/2014 March/2014 April/2014 May/2014 June/2014 Phase/Focus Instantiation of Project Review Literature Description Establish and instantiate the project along with all organizational constructs and administrative duties and responsibilities. Review the Literature on suicide keywords to better understand structure of queries when system is developed. Perform a content analysis of suicide notes and identify commonalities, patterns, regular expressions, etc Develop utility to collect the data from Facebook and Twitter feeds. The query parameters are the results form the literature review.
Collect data from the query results. Analyze the content of the searches (a subset of the data). Perform interpreter reliability tests for coder reliability. Modify utility based upon interpreter findings. Validated the utility by reanalyzing (a larger subset), treating utility with inter rater reliability metrics as well. Develop a validated metric to code profiles for suicide based on current literature
Identify and query profiles. Three coders analyze subset of profiles based on the metric from the previous phase. Develop a portion of utility to assess these automatically. Inter rater reliability with machine to test validity and reliability.
Identify offline methods and how those could translate to online methods of interventions. This includes literature reviews on what has been used as well as proof of concepts on what could potentially be done within the cyber-psychology domain.
I-5
Revised: 08/01/2011
Title
L-1
Revised: 08/01/2011
REFERENCES
Works Cited
American Foundation for Suicide Prevention. (n.d.). Facts and Figures. Retrieved November 27, 2011, from American Foundation for Suicide Prevention: http://www.afsp.org/index.cfm?page_id=04ea1254-bd31-1fa3-c549d77e6ca6aa37 Beck, A., Steer, R., & Ranieri, W. (1988). Scale for Suicide Ideation: Psychometric Properties of a SelfReport Version. Journal of Clinical Psychology . Bollen, J., Mao, H., & Zeng, X.-J. (2010). Twitter mood predicts the stock market. Journal of Computational Science . Canales, R. (2011). Suicide Risk Assessment: Early recognition of risk factors is key to prevention. Advance for Nurses . CVP Marketing Group. (2011, March 7). Facebook Demographics Revisited. Retrieved November 29, 2011, from Facebook: https://www.facebook.com/note.php?note_id=197149076992338 Golbeck, J., Robles, C., & Turner, K. (2011). Predicting personality with social media. Proceedings of the 2011 annual conference extended abstracts on Human factors in computing systems . Gottschalk, L., & Glester, G. (1960). An analysis of the verbal content of suicide notes. British Journal of Medical Psychology , 195-204. King, K., & Smith, J. (2000). Project SOAR: A Training Program to Increase School Counselors' Knowledge and Confidence Regarding Suicide Prevention and Intervention. Journal of School Health , 402-407. Marimoto, S., & Friedland, L. (2010). The Lifeworld of Youth in the Information Society. Youth and Society , 549-567. Moreno, M., Jelenchick, L., Egan, K., Cox, E., Young, H., Gannon, K., et al. (2011). Feeling bad on Facebook: depression disclosures by college students on a social networking site. Depression and Anxiety , 447-455. National Institute of Health. (2011, March 7). Suicide and SUicidal Behavior. Retrieved 12 1, 2011, from Medline Plus: http://www.nlm.nih.gov/medlineplus/ency/article/001554.htm Osgood, C., & Walker, E. (1959). Motivation and language behavior: A content analysis of suicide notes. The Journal of Abnormal and Social Psychology , 58-67. Quin, P., Agerbo, E., & Mortensen, P. B. (2003). Suicide risk in relation to socioeconomic, demographic, psychiatric, and familial factors: A national register-based study of all suicides in Denmark, 19811997. The American Journal of Psychiatry , 765-722. Ruder, T., Hatch, G., Ampanozi, G., Thali, M., & Fischer, N. (2011). Suicide Announcement on Facebook. Crisis: The Journal of Crisis Intervention and Suicide Prevention , 280-282. Russell, M. (2011). Mining the Social Web. Sebastopol: O'Reilly Media Inc. Tierney, R. (1994). Suicide Intervention training evaluation: A preliminary report. Crisis: The Journal of Crisis Intervention and Suicide Prevention , 69-76. Tuckman, J., Kleiner, R., & Lavell, M. (1959). Emotional content of suicide notes. The American Journal of Psychiatry , 59-63.
L-2
Revised: 08/01/2011